path: root/doc/gdbm.texi

\input texinfo  @c -*- Texinfo -*-
@comment %**start of header (This is for running Texinfo on a region.)
@setfilename gdbm.info
@include version.texi
@settitle GDBM manual

@ifinfo
@dircategory Programming & development tools
@direntry
* GDBM: (gdbm).                  The GNU database manager.
* gdbm_dump: (gdbm) gdbm_dump.   Dump the GDBM database into a flat file.
* gdbm_load: (gdbm) gdbm_load.   Load the database from a flat file.
* gdbmtool: (gdbm) gdbmtool.     Examine and modify a GDBM database.
@end direntry
@end ifinfo

@c @setchapternewpage odd
@comment %**end of header (This is for running Texinfo on a region.)

@c Use @kwindex for keywords
@defcodeindex kw
@syncodeindex kw cp
@c Use @flindex for files
@defcodeindex fl
@syncodeindex fl cp
@c Use @prindex for programs
@defcodeindex pr
@syncodeindex pr cp

@c Merge all indices into a single one
@syncodeindex fn cp
@syncodeindex vr cp
@syncodeindex ky cp
@syncodeindex pg cp
@syncodeindex tp cp

@iftex
@finalout
@end iftex

@copying
Published by the Free Software Foundation,
51 Franklin Street, Fifth Floor
Boston, MA 02110-1301, USA

Copyright @copyright{} 1989--2023 Free Software Foundation, Inc.

Permission is granted to copy, distribute and/or modify this document
under the terms of the GNU Free Documentation License, Version 1.3 or
any later version published by the Free Software Foundation; with no
Invariant Sections, no Front-Cover, and no Back-Cover texts.
A copy of the license is included in the section entitled ``GNU Free
Documentation License.''
@end copying

@titlepage
@sp 6
@center @titlefont{GNU dbm}
@sp 2
@center A Database Manager
@sp 2
@center by Philip A.  Nelson, Jason Downs and Sergey Poznyakoff
@sp 4
@center Manual by Pierre Gaumond, Philip A.  Nelson, Jason Downs,
@center Sergey Poznyakoff, and Terence Kelly
@sp 1
@center Edition @value{EDITION}
@sp 1
@center for GNU @command{dbm}, Version @value{VERSION}
@page
@vskip 0pt plus 1filll
@insertcopying
@end titlepage

@ifnothtml
@page
@summarycontents
@page
@end ifnothtml
@contents

@ifnottex
@node Top
@top The GNU database manager

GNU @command{dbm} (@command{GDBM}) is a library of functions
implementing a hashed database on a disk file.  This manual documents
GNU @command{dbm} Version @value{VERSION}.  The software was
originally written by Philip A.@: Nelson.  This document was
originally written by Pierre Gaumond from texts written by Phil.
@end ifnottex

@menu
* Copying::                    Your rights.
* Intro::                      Introduction to GNU dbm.

Functions:

* Open::                       Opening the database.
* Close::                      Closing the database.
* Count::                      Counting records in the database.
* Store::                      Inserting and replacing records in the database.
* Fetch::                      Searching records in the database.
* Delete::                     Removing records from the database.
* Sequential::                 Sequential access to records.
* Reorganization::             Database reorganization.
* Sync::                       Insure all writes to disk have competed.
* Database format::            GDBM database formats.
* Flat files::                 Export and import to Flat file format.
* Errors::                     Error handling.
* Database consistency::       Structural and logical consistency.
* Recovery::                   Recovery from fatal errors.
* Crash Tolerance::            Ensuring recovery to a consistent state.
* Options::                    Setting internal options.
* Locking::                    File locking.
* Variables::                  Useful global variables.
* Additional functions::       Functions for verifying internal structures.
* Error codes::                Error codes returned by GDBM calls.
* Compatibility::              Compatibility with UNIX dbm and ndbm.

Programs

* gdbmtool::                   Examine and modify a GDBM database.
* gdbm_dump::                  Dump the database into a flat file.
* gdbm_load::                  Load the database from a flat file.
* Exit codes::                 Exit codes returned by GDBM utilities.

Other topics:

* Bugs::                       Problems and bugs.
* Resources::                  Additional resources,

* GNU Free Documentation License::      Document license.
* Index::                       Index
@ifset WEBDOC
@ifhtml
* This Manual in Other Formats::
@end ifhtml
@end ifset

@detailmenu
 --- The Detailed Node Listing ---

Compatibility with standard @command{dbm} and @command{ndbm}

* ndbm::  NDBM interface functions.
* dbm::   DBM interface functions.

Examine and modify a GDBM database

* invocation::
* shell::

gdbmtool interactive mode

* variables::      shell variables.
* commands::       shell commands.
* definitions::    how to define structured data.
* startup files::

@end detailmenu
@end menu

@node Copying
@chapter Copying Conditions
This library is @dfn{free}; this means that everyone is free to use
it and free to redistribute it on a free basis.  GNU @command{dbm}
(@command{GDBM}) is not in the public domain; it is copyrighted and there
are restrictions on its distribution, but these restrictions are
designed to permit everything that a good cooperating citizen would want
to do.  What is not allowed is to try to prevent others from further
sharing any version of @command{GDBM} that they might get from
you.

  Specifically, we want to make sure that you have the right to give
away copies of @command{GDBM}, that you receive
source code or else can get it if you want it, that you can change these
functions or use pieces of them in new free programs, and that you know
you can do these things.

  To make sure that everyone has such rights, we have to forbid you to
deprive anyone else of these rights.  For example, if you distribute
copies of @command{GDBM}, you must give the recipients all
the rights that you have.  You must make sure that they, too, receive or
can get the source code.  And you must tell them their rights.

  Also, for our own protection, we must make certain that everyone finds
out that there is no warranty for anything in the @command{GDBM} distribution.
If these functions are modified by someone else and passed on, we want
their recipients to know that what they have is not what we distributed,
so that any problems introduced by others will not reflect on our
reputation.

@command{GDBM} is currently distributed under the terms of the GNU General
Public License, Version 3.  (@emph{NOT} under the GNU General Library
Public License.)  A copy the GNU General Public License is included with
the distribution of @command{GDBM}.

@node Intro
@chapter Introduction to GNU @command{dbm}

GNU @command{dbm} (@command{GDBM}) is a library of database functions that use
extensible hashing and work similar to the standard UNIX @command{dbm}
functions.  These routines are provided to a programmer needing to
create and manipulate a hashed database. (@command{GDBM} is @emph{NOT} a
complete database package for an end user.)

@tpindex datum
The basic use of @command{GDBM} is to store key/data pairs in a data file.
Each key must be unique and each key is paired with only one data item.
The keys can not be directly accessed in sorted order.  The basic unit
of data in @command{GDBM} is the structure:

@example
typedef struct
@{
   char *dptr;
   int  dsize;
@} datum;
@end example

This structure allows for arbitrary sized keys and data items.  In
particular, zero-length keys or data (@code{dsize = 0}) are allowed.
However, the @code{dptr} field is required to point to a valid memory
location.  In other words, @code{dptr} cannot be NULL.  Note also that
its type is @code{char *} for purely historic reasons.  You can use
any C data type (either scalar or aggregate) both as for key and for
data.

The key/data pairs are stored in a @command{GDBM} disk file, called a
@dfn{gdbm database}.  An application must open a @command{GDBM} database
to be able manipulate the keys and data contained in it.
@command{GDBM} allows an application to have multiple databases open at the
same time.  When an application opens a @command{GDBM} database, it is
designated as a @code{reader} or a @code{writer}.  A @command{GDBM}
database can be opened by at most one writer at a time.  However, many
readers may open the database simultaneously.  Readers and writers can
not open the @command{GDBM} database at the same time.

Speaking about @dfn{application} we usually mean a separate process.
However, it is entirely normal for a multi-thread program to operate
as a @command{GDBM} reader in one thread and writer in another, provided,
of course, that the two threads don't operate on the same database
simultaneously.

@flindex gdbm.h
To use the @command{GDBM} functions, the programmer must first include
the header file @file{gdbm.h}.

@tpindex GDBM_FILE
This file defines, among others, the @code{GDBM_FILE} data type, an
opaque pointer to the structure that represents the opened @command{GDBM}
database.  To access the database, the programmer must first open it
using the @code{gdbm_open} function.  The function takes several
arguments, the name of the database file being one of them, and
returns a @code{GDBM_FILE} object on success.  This object is then
passed to other functions in order to manipulate the database.  When
the database is no longer needed, the programmer @dfn{closes} it using
the @code{gdbm_close} call.

These and other functions are discussed in detail in chapters that
follow.  Here we show an example illustrating the use of @command{GDBM}
to look up a key in the database.

@example
#include <stdio.h>
#include <string.h>
#include <gdbm.h>

int
main (int argc, char **argv)
@{
  GDBM_FILE gdbf;     /* Database file object pointer */
  datum key, content; /* Key and content data */
  int status = 0;     /* Exit status of the program: 0 - OK, 1 - key
                         not found, 2 - error. */

  /*
   * Validate arguments.
   */
  if (argc != 3)
    @{
      fprintf (stderr, "usage: %s DBFILE KEY\n", argv[0]);
      return 2;
    @}

  /*
   * Open the database.  The GDBM_READER flag indicates that we only
   * intend to read from it.
   */
  gdbf = gdbm_open (argv[1], 0, GDBM_READER, 0, NULL);
  if (gdbf == NULL)
    @{
      fprintf (stderr, "can't open database: %s\n",
               gdbm_strerror (gdbm_errno));
    @}

  /*
   * Prepare the lookup key.  Notice, that the terminating \0 character
   * is not counted in the dsize computation.
   */
  key.dptr = argv[2];
  key.dsize = strlen (argv[2]);

  /*
   * Look up the key in the database.
   */
  content = gdbm_fetch (gdbf, key);

  /*
   * Analyze the return.
   */
  if (content.dptr != NULL)
    @{
      /*
       * The key is found.  Print the content on the stdout and
       * indicate success.
       */
      fwrite (content.dptr, content.dsize, 1, stdout);
      putchar ('\n');
      status = 0;
    @}
  else if (gdbm_errno == GDBM_ITEM_NOT_FOUND)
    @{
      /*
       * There is no such key in the database.
       */
      fprintf (stderr, "no such key\n");
      status = 1;
    @}
  else
    @{
      /*
       * An error occurred.
       */
      fprintf (stderr, "%s\n", gdbm_db_strerror (gdbf));
      status = 2;
    @}

  /*
   * Close the database and return.
   */
  gdbm_close (gdbf);
  return status;
@}
@end example

To compile this example, run

@example
cc -oexample example.c -lgdbm
@end example

To run it, you will need an example database.  The easiest way to
create it is by using the @command{gdbtool} program, which is part
of the @command{GDBM} package (@pxref{gdbmtool}):

@example
$ gdbmtool test.gdbm store foo bar
@end example

@noindent
This creates database file @file{test.gdbm} and stores a single record
in it.  The record's key is @samp{foo}, and the value is @samp{bar}.
Now you can run the example program to see how it works:

@example
$ ./example test.gdbm foo
bar
$ ./example test.gdbm baz
no such key
@end example

@node Open
@chapter Opening the database

@cindex opening the database
@cindex database, opening or creating
@deftypefn {gdbm interface} GDBM_FILE gdbm_open (const char *@var{name}, int @var{block_size}, @
  int @var{flags}, int @var{mode}, void (*@var{fatal_func})(const char *))
Opens or creates a @command{GDBM} database file.

The arguments are:

@table @var
@item name
The name of the file (the complete name, @command{GDBM} does not append any
characters to this name).

@item block_size
This parameter is used only when @code{gdbm_open} has to create a new
database file and represents the size of a single transfer from disk to
memory.  If its value is less than 512, the file system block
size is used instead.  The size is adjusted so that the block can hold
exact number of directory entries, so that the effective block size
can be slightly greater than requested.  However, if the
@code{GDBM_BSEXACT} flag is set and the size needs to be adjusted, the
function will return with error status, setting the @code{gdbm_errno}
variable to @code{GDBM_BLOCK_SIZE_ERROR}.

@item flags
@kwindex GDBM_READER
@kwindex GDBM_WRITER
@kwindex GDBM_WRCREAT
@kwindex GDBM_NEWDB
If @code{flags} is set to @code{GDBM_READER}, the user wants to just read the
database and any call to @code{gdbm_store} or @code{gdbm_delete} will fail.
Many readers can access the database at the same time.  If @code{flags} is
set to @code{GDBM_WRITER}, the user wants both read and write access
to the database and requires exclusive access.  If @code{flags} is set
to @code{GDBM_WRCREAT}, the user wants both read and write access to
the database and wants it created if it does not already exist.  If
@code{flags} is set to @code{GDBM_NEWDB}, the user want a new database
created, regardless of whether one existed, and wants read and write
access to the new database.  If an existing database file is opened with
the @code{GDBM_NEWDB} flag, the existing data are destroyed, and an
empty database structure is created in its place.

The following constants may also be logically or'd into the database
flags:

@defvr {gdbm_open flag} GDBM_CLOEXEC
@cindex close-on-exec
Set the close-on-exec flag on the database file descriptor.  The
@code{libc} must support the @code{O_CLOEXEC} flag
(@pxref{O_CLOEXEC,,,open(2),open(2) man page}).
@end defvr

@defvr {gdbm_open flag} GDBM_NOLOCK
Don't lock the database file.  Use this flag if you intend to do
locking separately.  @xref{Locking}.
@end defvr

@defvr {gdbm_open flag} GDBM_NOMMAP
Disable memory mapping mechanism.  Note, that this degrades performance.
@end defvr

@defvr {gdbm_open flag} GDBM_PREREAD
When mapping @command{GDBM} file to memory, read its contents immediately,
instead of when needed (@dfn{prefault reading}).  This can be
advantageous if you open a @emph{read-only} database and are going to
do a lot of look-ups on it.  In this case entire database will be
pre-read and look-ups will operate on an in-memory copy.  In 
contrast, @code{GDBM_PREREAD} should not be used if you open a
database (even in read-only mode) only to do a couple of look-ups.
Finally, never use @code{GDBM_PREREAD} when opening a database for
updates, especially for inserts: this will degrade performance.

This flag has no effect if @code{GDBM_NOMMAP} is given, or if the
operating system does not support prefault reading.  It is known
to work on Linux and FreeBSD kernels.
@end defvr

@defvr {gdbm_open flag} GDBM_XVERIFY
Enable additional consistency checks.  With this flag, eventual
corruptions of the database are discovered when opening it, instead of
when a corrupted structure is read during normal operation.  However,
on large databases, it can slow down the opening process.

@xref{Additional functions}.
@end defvr

The following additional flags are valid when the database is opened
for writing (i.e. together with @code{GDBM_WRITER},
@code{GDBM_WRCREAT}, or @code{GDBM_NEWDB}):

@defvr {gdbm_open flag} GDBM_SYNC
Synchronize all database operations to disk immediately.  Notice, that
this option entails severe performance degradation and does not
necessarily ensure that the resulting database state is consistent.
In general, we discourage its use (@pxref{Sync}).
@xref{Crash Tolerance}, for a discussion of how to ensure database
consistency with minimal performance overhead. 
@end defvr

@defvr {gdbm_open flag} GDBM_FAST
A reverse of @code{GDBM_SYNC}.  Synchronize writes only when needed.
This is the default.  The flag is provided for compatibility with
previous versions of @command{GDBM}.
@end defvr

The following flags can be used together with @code{GDBM_NEWDB}.  They
also take effect when used with @code{GDBM_WRCREAT}, if the requested
database file doesn't exist:

@defvr {gdbm_open flag} GDBM_BSEXACT
If this flag is set and the requested @var{block_size} cannot be used
without adjustment, @code{gdbm_open} will refuse to create the
databases.  In this case it will set the @code{gdbm_errno}
variable to @code{GDBM_BLOCK_SIZE_ERROR} and return @code{NULL}.
@end defvr

@defvr {gdbm_open flag} GDBM_NUMSYNC
Useful only together with @code{GDBM_NEWDB}, this bit instructs
@code{gdbm_open} to create new database in @dfn{extended database
format}, a format best suitable for effective crash recovery.
@xref{Numsync}, for a detailed discussion of this format, and
@ref{Crash Tolerance}, for a discussion of crash recovery.
@end defvr

@item mode
File mode@footnote{@xref{chmod,,,chmod(2),chmod(2) man page},
and @xref{open,,open a file,open(2), open(2) man page}.},
which is used if the file is created.

@item fatal_func
This parameter is deprecated and must always be @code{NULL}.

Early versions of @command{GDBM} (prior to 1.13) lacked proper error
handling and would abort on any ``fatal'' error (such as out of memory
condition, disk write error, or the like).  In these versions,
@code{fatal_func} was provided as a hook, allowing the caller to do
proper cleanup before such abnormal exit.  As of version
@value{VERSION}, this functionality is deprecated, although still
supported for backward compatibility.
@end table

The return value, is the pointer needed by all other functions to
access that @command{GDBM} file.  If the return is the @code{NULL} pointer,
@code{gdbm_open} was not successful.  The errors can be found in
@code{gdbm_errno} variable (@pxref{Variables, gdbm_errno}).  Available
error codes are discussed in @ref{Error codes}.

In all of the following calls, the parameter @var{dbf} refers to the pointer
returned from @code{gdbm_open} (or @code{gdbm_fd_open}, described below).
@end deftypefn

@anchor{gdbm_fd_open}
@deftypefn {gdbm interface} GDBM_FILE gdbm_fd_open (int @var{fd},@
  const char *@var{name}, int @var{block_size}, @
  int @var{flags}, void (*@var{fatal_func})(const char *))

Alternative function for opening a @command{GDBM} database.  The @var{fd}
argument is the file descriptor of the database file obtained by a
call to @code{open}(2), @code{creat}(2) or similar functions.  The
descriptor is not dup'ed, and will be closed when the returned
@code{GDBM_FILE} is closed.  Use @code{dup}(2) if that is not
desirable.

In case of error, the function behaves like @code{gdbm_open} and
@emph{does not close} @var{fd}.  This can be altered by the following
value passed in the @var{flags} argument:

@defvr {gdbm_open flag} GDBM_CLOERROR
Close @var{fd} before exiting on error.
@end defvr
@end deftypefn

@deftypefn {gdbm interface} int gdbm_copy_meta (GDBM_FILE @var{dst},@
 GDBM_FILE @var{src})
Copy file ownership and mode from @var{src} to @var{dst}.
@end deftypefn

@node Close
@chapter Closing the database
@cindex closing database
@cindex database, closing

It is important that every file opened is also closed.  This is needed
to properly update its disk structure and maintain a consistent
locking state on the file.

@deftypefn {gdbm interface} int gdbm_close (GDBM_FILE @var{dbf})
This function closes the @command{GDBM} file and frees all memory
associated with it.  The parameter is:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@end table

@code{Gdbm_close} returns 0 on success. On error, it sets
@code{gdbm_errno} and system @code{errno} variables to the codes
describing the error and returns -1.
@end deftypefn

@node Count
@chapter Number of Records
@cindex number of records
@deftypefn {gdbm interface} int gdbm_count (GDBM_FILE @var{dbf}, @
  gdbm_count_t *@var{pcount})
Counts the number of records in the database @var{dbf}.  On success,
stores it in the memory location pointed to by @var{pcount} and returns
0.  On error, sets @code{gdbm_errno} (if relevant, also @code{errno})
and returns -1.
@end deftypefn

@deftypefn {gdbm interface} int gdbm_bucket_count (GDBM_FILE @var{dbf}, @
  size_t *@var{pcount})
Counts the number of buckets in the database @var{dbf}.  On success,
stores it in the memory location pointed to by @var{pcount} and return
0.  On error, sets @code{gdbm_errno} (if relevant, also @code{errno})
and returns -1.
@end deftypefn

@node Store
@chapter Inserting and replacing records in the database
@cindex storing records
@cindex records, storing

@deftypefn {gdbm interface} int gdbm_store (GDBM_FILE @var{dbf}, datum @var{key}, @
           datum @var{content}, int @var{flag})
The function @code{gdbm_store} inserts or replaces records in the database.

The parameters are:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@item key
The search key.
@item content
The data to be associated with the key.
@item flag
@kwindex GDBM_REPLACE
@kwindex GDBM_INSERT
Defines the action to take when the key is already in the database.  The value
@code{GDBM_REPLACE} asks that the old data be replaced by the new
@var{content}.  The value @code{GDBM_INSERT} asks that an error be
returned and no action taken if the @var{key} already exists.
@end table

This function can return the following values:

@table @asis
@item 0
Success.  The value of @var{content} is keyed by @var{key} in the
database.

@item -1
An error occurred which prevented the item from being stored in the
database.   Examine the @code{gdbm_errno} variable to determine the
actual cause of the error.

@item +1
The item was not stored because the argument @var{flag} was
@code{GDBM_INSERT} and the @var{key} was already in the database.
The @code{gdbm_errno} variable is set to @code{GDBM_CANNOT_REPLACE}.
@end table

If the function returns -1, @code{gdbm_errno} can have the following
values:

@table @code
@item GDBM_READER_CANT_STORE
Database was open in read-only mode, i.e. with the @code{GDBM_READER}
flag.  @xref{Open}.

@item GDBM_MALFORMED_DATA
Either @var{key} or @var{content} had their @code{dptr} field set to
@code{NULL}.

It is OK to have a @dfn{zero-length} key or content, i.e. a datum with
@code{dsize} set to 0, but the @code{dptr} field must always be a
non-NULL value.

@item GDBM_BAD_HASH_TABLE
Database hash table is malformed.  This usually means that some error
in the application or the library caused memory overrun.  The database
is marked as needing recovery.  All further calls on this database
will return with @code{gdbm_error} set to @code{GDBM_NEED_RECOVERY}.
@xref{Recovery}, for a discussion of database recovery process.

@item GDBM_BAD_DIR_ENTRY
Database directory entry is corrupted.  The database is marked as
needing recovery.  @xref{Recovery}.

@item GDBM_BAD_BUCKET
Database bucket is corrupted.  The database is marked as
needing recovery.  @xref{Recovery}.

@item GDBM_BAD_AVAIL
Database available storage index is corrupted.  The database is marked as
needing recovery.  @xref{Recovery}.

@item GDBM_FILE_SEEK_ERROR
A seek error occurred on the underlying disk file.  Examine the system
@code{errno} variable for more detail.
@end table

@end deftypefn

If you store data for a @var{key} that is already in the data base,
@command{GDBM} replaces the old data with the new data if called with
@code{GDBM_REPLACE}.  You do not get two data items for the same
@code{key} and you do not get an error from @code{gdbm_store}.

The size of datum in @command{GDBM} is restricted only by the maximum
value for an object of type @code{int} (type of the @code{dsize} member of
@code{datum}).

@node Fetch
@chapter Searching for records in the database
@cindex fetching records
@cindex looking up records
@cindex record, fetching

@deftypefn {gdbm interface} datum gdbm_fetch (GDBM_FILE @var{dbf}, datum @var{key})
Looks up a given @var{key} and returns the information associated with it.
The @code{dptr} field in the structure that is returned points to a
memory block allocated by @code{malloc}.  It is the caller's
responsibility to free it when no longer needed.

If the @code{dptr} is @code{NULL}, inspect the value of the
@code{gdbm_errno} variable (@pxref{Variables,gdbm_errno}).  If it is
@code{GDBM_ITEM_NOT_FOUND}, no data was found.  Any other value means an
error occurred.  Use @code{gdbm_strerror} function to convert
@code{gdbm_errno} to a human-readable string.

The parameters are:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@item key
The search key.
@end table
@end deftypefn

An example of using this function:

@example
content = gdbm_fetch (dbf, key);
if (content.dptr == NULL)
  @{
    if (gdbm_errno == GDBM_ITEM_NOT_FOUND)
      fprintf(stderr, "key not found\n");
    else
      fprintf(stderr, "error: %s\n", gdbm_db_strerror (dbf));
  @}
else
  @{
    /* do something with content.dptr */
  @}
@end example

@cindex records, testing existence
You may also search for a particular key without retrieving it:

@deftypefn {gdbm interface} int gdbm_exists (GDBM_FILE @var{dbf}, datum @var{key})
Checks whether the @var{key} exists in the database @var{dbf}.

If @var{key} is found, returns @code{true} (@code{1}).  If it is not
found, returns @code{false} (@code{0}) and sets @code{gdbm_errno} to
@code{GDBM_NO_ERROR} (@code{0}).

On error, returns @code{0} and sets @code{gdbm_errno} to a
non-@code{0} error code.

The parameters are:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@item key
The search key.
@end table
@end deftypefn

@node Delete
@chapter Removing records from the database
@cindex deleting records
@cindex record, deleting

To remove some data from the database, use the @code{gdbm_delete}
function.

@deftypefn {gdbm interface} int gdbm_delete (GDBM_FILE @var{dbf}, datum @var{key})
Deletes the data associated with the given @var{key}, if it exists in
the database @var{dbf}.

The parameters are:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@item datum key
The search key.
@end table

The function returns @code{-1} if the item is not present or if an
error is encountered.  Examine the @code{gdbm_errno} variable or
the return from @code{gdbm_last_errno (@var{dbf})} to know the reason.

The return of @code{0} marks a successful delete.
@end deftypefn

@node Sequential
@chapter Sequential access to records
@cindex sequential access
@cindex iterating over records
@cindex records, iterating over

The next two functions allow for accessing all items in the database.  This
access is not @code{key} sequential, but it is guaranteed to visit every
@code{key} in the database once.  The order has to do with the hash values.
@code{gdbm_firstkey} starts the visit of all keys in the database.
@code{gdbm_nextkey} finds and reads the next entry in the hash structure for
@code{dbf}.

@deftypefn {gdbm interface} datum gdbm_firstkey (GDBM_FILE @var{dbf})
Initiate sequential access to the database @var{dbf}.  The returned
value is the first key accessed in the database.  If the @code{dptr}
field in the returned datum is @code{NULL}, inspect the
@code{gdbm_errno} variable (@pxref{Variables, gdbm_errno}).  The value
of @code{GDBM_ITEM_NOT_FOUND} means that the database contains no
data.  Other value means an error occurred.

On success, @code{dptr} points to a memory block obtained from
@code{malloc}, which holds the key value.  The caller is responsible
for freeing this memory block when no longer needed.
@end deftypefn

@deftypefn {gdbm interface} datum gdbm_nextkey (GDBM_FILE @var{dbf}, datum @var{prev})
This function continues iteration over the keys in @var{dbf},
initiated by @code{gdbm_firstkey}.  The parameter @var{prev} holds the
value returned from a previous call to @code{gdbm_nextkey} or
@code{gdbm_firstkey}.

The function returns next key from the database.  If the @code{dptr}
field in the returned datum is @code{NULL} inspect the
@code{gdbm_errno} variable (@pxref{Variables, gdbm_errno}).  The value
of @code{GDBM_ITEM_NOT_FOUND} means that all keys in the database
has been visited.  Any other value means an error occurred.

Otherwise, @code{dptr} points to a memory block obtained from
@code{malloc}, which holds the key value.  The caller is responsible
for freeing this memory block when no longer needed.
@end deftypefn

@cindex iteration loop
These functions are intended to visit the database in read-only algorithms,
for instance, to validate the database or similar operations.  The
usual algorithm for sequential access is:

@example
@group
   key = gdbm_firstkey (dbf);
   while (key.dptr)
     @{
        datum nextkey;

        /* do something with the key */
        ...

        /* Obtain the next key */
        nextkey = gdbm_nextkey (dbf, key);
        /* Reclaim the memory used by the key */
        free (key.dptr);
        /* Use nextkey in the next iteration. */
        key = nextkey;
     @}
@end group
@end example

@cindex iteration and @code{gdbm_delete}
@cindex deletion in iteration loops
@cindex @code{gdbm_delete} and sequential access
Don't use @code{gdbm_delete} or @code{gdbm_store} in
such a loop.  File visiting is based on a @dfn{hash table}.  The
@code{gdbm_delete} function re-arranges the hash table to make sure
that any collisions in the table do not leave some item
@dfn{un-findable}.  The original key order is @emph{not} guaranteed to
remain unchanged in all instances.  So it is possible that some key
will not be visited or will be visited twice, if a loop like the
following is executed:

@example
@group
   key = gdbm_firstkey (dbf);
   while (key.dptr)
     @{
        datum nextkey;
        if (some condition)
          @{
             gdbm_delete (dbf, key);
          @}
         nextkey = gdbm_nextkey (dbf, key);
         free (key.dptr);
         key = nextkey;
      @}
@end group
@end example

@node Reorganization
@chapter Database reorganization
@cindex database reorganization
@cindex reorganization, database

The following function should be used very seldom.

@deftypefn {gdbm interface} int gdbm_reorganize (GDBM_FILE @var{dbf})
Reorganizes the database.

The parameter is:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@end table
@end deftypefn

If you have had a lot of deletions and would like to shrink the space
used by the @command{GDBM} file, this function will reorganize the database.
This results, in particular, in shortening the length of a @command{GDBM}
file by removing the space occupied by deleted records.

This reorganization requires creating a new file and inserting all the elements
in the old file @var{dbf} into the new file.  The new file is then renamed to
the same name as the old file and @var{dbf} is updated to contain all the
correct information about the new file.  If an error is detected, the return
value is negative.  The value zero is returned after a successful
reorganization.

@node Sync
@chapter Database Synchronization
@cindex database synchronization
@cindex synchronization, database

Normally, @command{GDBM} functions don't flush changed data to the
disk immediately after a change.  This allows for faster writing of
databases at the risk of having a corrupted database if the
application terminates in an abnormal fashion.  The following function
allows the programmer to make sure the disk version of the database
has been completely updated with all changes to the current time.

@deftypefn {gdbm interface} int gdbm_sync (GDBM_FILE @var{dbf})
Synchronizes the changes in @var{dbf} with its disk file.  The
parameter is a pointer returned by @code{gdbm_open}.

This function would usually be called after a complete set of changes
have been made to the database and before some long waiting time.
This set of changes should preserve application-level invariants.  In
other words, call @code{gdbm_sync} only when the database is in a
consistent state with regard to the application logic, a state from
which you are willing and able to recover.  You can think about all
database operations between two consecutive @code{gdbm_sync} calls as
constituting a single @dfn{transaction}.  @xref{Synchronizing the
Database}, for a detailed discussion about how to properly select
the synchronization points.

The @code{gdbm_close} function automatically calls the equivalent of
@code{gdbm_sync} so no call is needed if the database is to be closed
immediately after the set of changes have been made.

@code{Gdbm_sync} returns 0 on success. On error, it sets
@code{gdbm_errno} and system @code{errno} variables to the codes
describing the error and returns -1.
@end deftypefn

@kwindex GDBM_SYNC
Opening the database with @code{GDBM_SYNC} flag ensures that
@code{gdbm_sync} function will be called after each change, thereby
flushing the changes to disk immediately.  You are advised against
using this flag, however, because it incurs a severe performance
penalty, while giving only a moderate guarantee that the
@emph{structural} consistency of the database will be preserved in case
of failure, and that only unless the failure occurs while being in the
@code{fsync} call.  For the ways to ensure proper @emph{logical} consistency
of the database, see @ref{Crash Tolerance}.

@node Database format
@chapter Changing database format
As of version @value{VERSION}, @command{GDBM} supports databases in
two formats: @dfn{standard} and @dfn{extended}.  The standard format
is used most often.  The @dfn{extended} database format is used to
provide additional crash resistance (@pxref{Crash Tolerance}).

Depending on the value of the @var{flags} parameter in a call to
@code{gdbm_open} (@pxref{Open, GDBM_NUMSYNC}), a database can be
created in either format.

The format of an existing database can be changed using the
@code{gdbm_convert} function:

@deftypefn {gdbm interface} int gdbm_convert (GDBM_FILE @var{dbf}, @
  int @var{flag})
Changes the format of the database file @var{dbf}.  Allowed values for
@var{flag} are:

@table @code
@item 0
Convert database to the standard format.

@kwindex GDBM_NUMSYNC
@item GDBM_NUMSYNC
Convert database to the extended @dfn{numsync} format (@pxref{Numsync}).
@end table

On success, the function returns 0.  In this case, it should be
followed by a call to @code{gdbm_sync} (@pxref{Sync}) or
@code{gdbm_close} (@pxref{Close}) to ensure the changes are written to
the disk.

On error, returns -1 and sets the @code{gdbm_errno} variable
(@pxref{Variables, gdbm_errno}).

If the database is already in the requested format, the function
returns success (0) without doing anything.
@end deftypefn

@node Flat files
@chapter Export and Import
@cindex Flat file format
@cindex export
@cindex import
@command{GDBM} databases can be converted into so-called @dfn{flat
format} files.  Such files cannot be used for searching, their sole
purpose is to keep the data from the database for restoring it when
the need arrives.  There are two flat file formats, which differ in
the way they represent the data and in the amount of meta-information
stored.  Both formats can be used, for example, to migrate between
the different versions of @command{GDBM} databases.  Generally speaking,
flat files are safe to send over the network, and can be used to
recreate the database on another machine.  The recreated database is
guaranteed to have the same format and contain the same set of
key/value pairs as the database from which the flat file was created.
However, it will not constitute a byte-to-byte equivalent of the latter.
Various internal structures in the database can differ.  In
particular, ordering of key/value pairs can be different and the table
of available file space will most probably differ, too.  For databases
in extended format, the @code{numsync} counter will be reset to 0
(@pxref{Numsync}).  These details are not visible to the application
programmer, and are mentioned here only for completeness sake.

The fact that the restored database contains the same set of key/value
pairs does not necessarily mean, however, that it can be used in the
same way as the original one.  For example, if the original database
contained non-@acronym{ASCII} data (e.g.@: @acronym{C} structures,
integers etc.), the recreated database can be of any use only if the
target machine has the same integer size and byte ordering as the
source one and if its @acronym{C} compiler uses the same packing
conventions as the one which generated @acronym{C} which populated the
original database.  In general, such binary databases are not portable
between machines, unless you follow some stringent rules on what data
is written to them and how it is interpreted.

@command{GDBM} version @value{VERSION} supports two flat file formats.  The
@dfn{binary} flat file format was first implemented in version
1.9.1.  This format stores only key/data pairs, it does not keep
information about the database file itself.  As its name implies,
files in this format are binary files.  This format is supported for
backward compatibility.

The @dfn{ascii} flat file format encodes all data in Base64 and stores
not only key/data pairs, but also the original database file metadata,
such as file name, mode and ownership.  Files in this format can be
sent without additional encapsulation over transmission channels that
normally allow only ASCII data, such as, e.g.@: SMTP.  Due to additional
metadata they allow for restoring an exact copy of the database,
including file ownership and privileges, which is especially important
if the database in question contained some security-related data.

We call a process of creating a flat file from a database
@dfn{exporting} or @dfn{dumping} this database.  The reverse process,
creating the database from a flat file is called @dfn{importing} or
@dfn{loading} the database.

@deftypefn {gdbm interface} int gdbm_dump (GDBM_FILE @var{dbf}, @
    const char *@var{filename}, int @var{format}, @
    int @var{open_flags}, int @var{mode})
Dumps the database file to the named file in requested format.
Arguments are:

@table @var
@item dbf
A pointer to the source database, returned by a prior call to
@code{gdbm_open}.

@item filename
Name of the dump file.

@item format
Output file format.  Allowed values are: @code{GDBM_DUMP_FMT_BINARY} to
create a binary dump and @code{GDBM_DUMP_FMT_ASCII} to create an ASCII
dump file.

@item open_flags
How to create the output file.  If @var{flag} is @code{GDBM_WRCREAT}
the file will be created if it does not exist.  If it does exist,
the @code{gdbm_dump} will fail.

If @var{flag} is @code{GDBM_NEWDB}, the function will create a new
output file, replacing it if it already exists.

@item mode
The permissions to use when creating the output file (@pxref{open,,open a file,open(2), open(2) man page}).
@end table
@end deftypefn

@anchor{gdbm_load function}
@deftypefn {gdbm interface} int gdbm_load (GDBM_FILE *@var{pdbf}, @
    const char *@var{filename}, int @var{flag}, @
    int @var{meta_mask}, @
    unsigned long *@var{errline})
Loads data from the dump file @var{filename} into the database pointed
to by @var{pdbf}.  The latter can point to @code{NULL}, in which case
the function will try to create a new database.  If it succeeds, the
function will return, in the memory location pointed to by @var{pdbf},
a pointer to the newly created database.  If the dump file carries no
information about the original database file name, the function will
set @code{gdbm_errno} to @code{GDBM_NO_DBNAME} and return
@code{-1}, indicating failure.

The @var{flag} has the same meaning as the @var{flag} argument
to the @code{gdbm_store} function (@pxref{Store}).

The @var{meta_mask} argument can be used to disable restoring certain
bits of file's meta-data from the information in the input dump file.
It is a binary OR of zero or more of the following:

@table @asis
@item GDBM_META_MASK_MODE
Do not restore file mode.

@item GDBM_META_MASK_OWNER
Do not restore file owner.
@end table

The function returns 0 upon successful completion or -1 on fatal
errors and 1 on mild (non-fatal) errors.

If a fatal error occurs, @code{gdbm_errno} will be set to one of the
following values:

@table @asis
@item GDBM_FILE_OPEN_ERROR
Input file (@var{filename}) cannot be opened.  The @code{errno}
variable can be used to get more detail about the failure.

@item GDBM_MALLOC_ERROR
Not enough memory to load data.

@item GDBM_FILE_READ_ERROR
Reading from @var{filename} failed.  The @code{errno} variable can be
used to get more detail about the failure.

@item GDBM_MALFORMED_DATA
@itemx GDBM_ILLEGAL_DATA
Input contained malformed data, i.e. it is not a valid @command{GDBM}
dump file.  This often means that the dump file got corrupted
during the transfer.

The @code{GDBM_ILLEGAL_DATA} is an alias for this error code,
maintained for backward compatibility.

@item GDBM_ITEM_NOT_FOUND
This error can occur only when the input file is in ASCII format.  It
indicates that the data part of the record about to be read lacked
length specification.  Application developers are advised to treat
this error equally as @code{GDBM_MALFORMED_DATA}.
@end table

Mild errors mean that the function was able to successfully load and
restore the data, but was unable to change the database file metadata
afterwards.  The table below lists possible values for @code{gdbm_errno}
in this case.  To get more detail, inspect the system @code{errno} variable.

@table @asis
@kwindex GDBM_ERR_FILE_OWNER
@item GDBM_ERR_FILE_OWNER
The function was unable to restore database file owner.

@kwindex GDBM_ERR_FILE_MODE
@item GDBM_ERR_FILE_MODE
The function was unable to restore database file mode (permission bits).
@end table

If an error occurs while loading data from an input file in ASCII
format, the number of line in which the error occurred will be stored
in the location pointed to by the @var{errline} parameter, unless it
is @code{NULL}.

If the line information is not available or applicable, @var{errline}
will be set to @code{0}.
@end deftypefn

@deftypefn {gdbm interface} int gdbm_dump_to_file (GDBM_FILE @var{dbf}, @
    FILE *@var{fp}, int @var{format})
This is an alternative entry point to @code{gdbm_dump} (which see).
Arguments are:

@table @var
@item dbf
A pointer to the source database, returned by a call to
@code{gdbm_open}.

@item fp
File to write the data to.

@item format
Format of the dump file.  See the @var{format} argument to the
@code{gdbm_dump} function.
@end table
@end deftypefn

@deftypefn {gdbm interface} int gdbm_load_from_file (GDBM_FILE *@var{pdbf}, @
    FILE *@var{fp}, int @var{replace}, int @var{meta_mask}, @
    unsigned long *@var{line})
This is an alternative entry point to @code{gdbm_load}.  It writes the
output to @var{fp} which must be a file open for writing.  The rest of
arguments is the same as for @code{gdbm_load} (excepting of course
@var{flag}, which is not needed in this case).
@end deftypefn

@deftypefn {gdbm interface} int gdbm_export (GDBM_FILE @var{dbf}, @
           const char *@var{exportfile}, int @var{flag}, int @var{mode})
This function is retained for compatibility with GDBM 1.10 and
earlier.  It dumps the database to a file in binary dump format and
is equivalent to

@example
gdbm_dump(@var{dbf}, @var{exportfile}, GDBM_DUMP_FMT_BINARY, @var{flag}, @var{mode})
@end example

@end deftypefn

@deftypefn {gdbm interface} int gdbm_export_to_file (GDBM_FILE @var{dbf}, FILE *@var{fp})
This is an alternative entry point to @code{gdbm_export}.  This
function writes to file @var{fp} a binary dump of the database @var{dbf}.
@end deftypefn

@deftypefn {gdbm interface} int gdbm_import (GDBM_FILE @var{dbf}, @
           const char *@var{importfile}, int @var{flag})
This function is retained for compatibility with @command{GDBM} 1.10 and
earlier.  It loads the file @var{importfile}, which must be a binary
flat file, into the database @var{dbf} and is equivalent to the
following construct:

@example
@var{dbf} = gdbm_open (@var{importfile}, 0,
                       @var{flag} == GDBM_REPLACE ?
                         GDBM_WRCREAT : GDBM_NEWDB,
                       0600, NULL);
gdbm_load (&@var{dbf}, @var{exportfile}, 0, @var{flag}, NULL)
@end example
@end deftypefn

@deftypefn {gdbm interface} int gdbm_import_from_file (GDBM_FILE @var{dbf}, @
      FILE *@var{fp}, int @var{flag})
An alternative entry point to @code{gdbm_import}.  Reads the binary
dump from the file @var{fp} and stores the key/value pairs to
@var{dbf}.  @xref{Store}, for a description of @var{flag}.

This function is equivalent to:

@example
@var{dbf} = gdbm_open (@var{importfile}, 0,
                       @var{flag} == GDBM_REPLACE ?
                         GDBM_WRCREAT : GDBM_NEWDB,
                       0600, NULL);
gdbm_load_from_file (@var{dbf}, @var{fp}, @var{flag}, 0, NULL);
@end example
@end deftypefn

@node Errors
@chapter Error handling
@cindex gdbm_errno
@cindex error strings
@cindex global error state
The global variable @code{gdbm_errno} (@pxref{Variables, gdbm_errno})
keeps the error code of the most recent error encountered by @command{GDBM}
functions.

To convert this code to human-readable string, use the following function:

@deftypefn {gdbm interface} {const char *} gdbm_strerror (gdbm_error @var{errno})
Converts @var{errno} (an integer value) into a human-readable
descriptive text.  Returns a pointer to a static string.  The caller
must not free the returned pointer or alter the string it points to.
@end deftypefn

Detailed information about the most recent error that occurred while
operating on a @command{GDBM} file is stored in the @code{GDBM_FILE} object
itself.  To retrieve it, the following functions are provided:

@cindex error code, most recent
@cindex most recent error code
@deftypefn {gdbm interface} {gdbm_error} gdbm_last_errno (GDBM_FILE @var{dbf})
Returns the code of the most recent error encountered when operating
on @var{dbf}.

When @code{gdbm_last_errno} called immediately after the failed
function, its return equals the value of the @code{gdbm_errno}
variable.  However, @code{gdbm_errno} can be changed if any
@command{GDBM} functions (operating on another databases) were called
afterwards, and @code{gdbm_last_errno} will always return the code of
the last error that occurred while working with @emph{that} database.
@end deftypefn

@deftypefn {gdbm interface} {int} gdbm_last_syserr (GDBM_FILE @var{dbf})
Returns the value of the system @code{errno} variable associated with
the most recent error.

Notice, that not all @command{GDBM} errors have an associated system error
code.  The following are the ones that have:

@itemize @bullet
@item GDBM_FILE_OPEN_ERROR
@item GDBM_FILE_WRITE_ERROR
@item GDBM_FILE_SEEK_ERROR
@item GDBM_FILE_READ_ERROR
@item GDBM_FILE_STAT_ERROR
@item GDBM_BACKUP_FAILED
@item GDBM_FILE_CLOSE_ERROR
@item GDBM_FILE_SYNC_ERROR
@item GDBM_FILE_TRUNCATE_ERROR
@item GDBM_ERR_SNAPSHOT_CLONE
@item GDBM_ERR_REALPATH
@item GDBM_ERR_USAGE
@end itemize

For other errors, @code{gdbm_last_syserr} will return 0.
@end deftypefn

@anchor{gdbm_check_syserr}
@deftypefn {gdbm interface} {int} gdbm_check_syserr (gdbm_errno @var{err})
Returns @code{1}, if the system @code{errno} value should be inspected
to get more info on the error described by @command{GDBM} error code
@var{err}.
@end deftypefn

To get a human-readable description of the recent error for a
particular database file, use the @code{gdbm_db_strerror} function:

@deftypefn {gdbm interface} {const char *} gdbm_db_strerror (GDBM_FILE @var{dbf})
Returns textual description of the most recent error encountered when
operating on the database @var{dbf}.  The resulting string is often
more informative than what would be returned by
@code{gdbm_strerror(gdbm_last_errno(@var{dbf}))}.  In particular, if
there is a system error associated with the recent failure, it will be
described as well.
@end deftypefn

@deftypefn {gdbm interface} void gdbm_clear_error (GDBM_FILE @var{dbf})
Clears the error state for the database @var{dbf}.  Normally, this
function is called upon the entry to any @command{GDBM} function.
@end deftypefn

Certain errors (such as write error when saving stored key) can leave
database file in inconsistent state (@pxref{Database consistency}).
When such a critical error occurs, the database file is marked as
needing recovery.  Subsequent calls to any @command{GDBM} functions for that
database file (except @code{gdbm_recover}), will return immediately
with @command{GDBM} error code @code{GDBM_NEED_RECOVERY}.  Additionally, the
following function can be used to check the state of the database file:

@deftypefn {gdbm interface} int gdbm_needs_recovery (GDBM_FILE @var{dbf})
Returns @code{1} if the database file @var{dbf} is in inconsistent
state and needs recovery.
@end deftypefn

To restore structural consistency of the database, use the
@code{gdbm_recover} function (@pxref{Recovery}).

Crash tolerance provides a better way of recovery, because it restores
both structural and logical consistency.  @xref{Crash Tolerance}, for
a detailed discussion,

@node Database consistency
@chapter Database consistency

@cindex consistency, database
In the chapters that follow we will cover different aspects of
@dfn{database consistency} and ways to maintain it.  Speaking
about consistency, it is important to distinguish between two different
aspects of it: structural and logical consistency.

@cindex structural consistency
@dfn{Structural consistency} means that all internal structures of the
database are in good order, contain valid data and are coherent with
one another.  Structural consistency means that the database is in
good shape @dfn{technically}, but it does not imply that the data it
contains are in any way meaningful.

@cindex logical consistency
@dfn{Logical consistency} means that the data stored in the
database are coherent with respect to the application logic.
Usually this implies that structural consistency is observed as well.

For as long as the program is free from memory management errors and
each opened database is properly closed before the program terminates,
structural consistency is maintained.  Maintaining logical consistency
is more complex task and its maintenance is entirely the
responsibility of the application programmer.  @xref{Crash Tolerance},
for a detailed discussion.

Both consistency aspects can suffer as a result of both application
errors that cause the program to terminate prematurely without properly
saving the database, and hardware errors, such as disk failures or
power outages.  When such situations occur, it becomes necessary to
@dfn{recover the database}.

In the next chapter we will discuss how to recover structural
consistency of a database.

@node Recovery
@chapter Recovering structural consistency

Certain errors (such as write error when saving stored key) can leave
database file in @dfn{structurally inconsistent state}.  When such a
critical error occurs, the database file is marked as needing
recovery.  Subsequent calls to any GDBM functions for that database
file (except @code{gdbm_recover}), will return immediately with
@command{GDBM} error code @code{GDBM_NEED_RECOVERY}.

To escape from this state and bring the database back to operational
state, use the following function:

@deftypefn {gdbm interface} int gdbm_recover (GDBM_FILE @var{dbf},@
 gdbm_recovery *@var{rcvr}, int @var{flags})
Check the database file @var{dbf} and fix eventual errors.  The
@var{rcvr} argument points to a structure that has @dfn{input
members}, providing additional information to alter the behavior of
@code{gdbm_recover}, and @dfn{output members}, which are used to return
additional statistics about the recovery process (@var{rcvr} can be
@code{NULL} if no such information is needed).

Each input member has a corresponding flag bit, which must be set in
@var{flags}, in order to instruct the function to use it.

The @code{gdbm_recover} type is defined as:

@example
typedef struct gdbm_recovery_s
@{
  /* Input members.
     These are initialized before call to gdbm_recover.
     The flags argument specifies which of them are initialized. */
  void (*errfun) (void *data, char const *fmt, ...);
  void *data;
  size_t max_failed_keys;
  size_t max_failed_buckets;
  size_t max_failures;

  /* Output members.
     The gdbm_recover function fills these before returning. */
  size_t recovered_keys;
  size_t recovered_buckets;
  size_t failed_keys;
  size_t failed_buckets;
  char *backup_name;
@} gdbm_recovery;
@end example

The @dfn{input members} modify the behavior of @code{gdbm_recover}:

@deftypecv {input member} gdbm_recovery void (*errfun) (void *@var{data},@
  char const *@var{fmt}, ...)
@kwindex GDBM_RCVR_ERRFUN
If the @code{GDBM_RCVR_ERRFUN} flag bit is set, @code{errfun} points
to a function that will be called upon each recoverable or non-fatal
error that occurred during the recovery.  The @code{data} field of
@code{gdbm_recovery} will be passed to it as its first argument.  The
@var{fmt} argument is a @code{printf}-like (@pxref{Format of the format string,,,printf(3), printf(3) man page}), format string.  The rest of
arguments supply parameters for that format.
@end deftypecv

@deftypecv {input member} gdbm_recovery {void *} data
Supplies first argument for the @code{errfun} invocations.
@end deftypecv

@deftypecv {input member} gdbm_recovery size_t max_failed_keys
@kwindex GDBM_RCVR_MAX_FAILED_KEYS
If @code{GDBM_RCVR_MAX_FAILED_KEYS} is set, this member sets the limit
on the number of keys that cannot be retrieved.  If the number of
failed keys becomes equal to @code{max_failed_keys}, recovery is
aborted and error is returned.
@end deftypecv

@deftypecv {input member} gdbm_recovery size_t max_failed_buckets
@kwindex GDBM_RCVR_MAX_FAILED_BUCKETS
If @code{GDBM_RCVR_MAX_FAILED_BUCKETS} is set, this member sets the limit
on the number of buckets that cannot be retrieved or that contain
bogus information.  If the number of failed buckets becomes equal to
@code{max_failed_buckets}, recovery is aborted and error is returned.
@end deftypecv

@deftypecv {output member} gdbm_recovery size_t max_failures
@kwindex GDBM_RCVR_MAX_FAILURES
If @code{GDBM_RCVR_MAX_FAILURES} is set, this member sets the limit
of failures that are tolerated during recovery.  If the number of
errors becomes equal to @code{max_failures}, recovery is aborted and
error is returned.
@end deftypecv

The following members are filled on output, upon successful return
from the function:

@deftypecv {output member} gdbm_recovery size_t recovered_keys
Number of recovered keys.
@end deftypecv

@deftypecv {output member} gdbm_recovery size_t recovered_buckets
Number of recovered buckets.
@end deftypecv

@deftypecv {output member} gdbm_recovery size_t failed_keys
Number of key/data pairs that could not be retrieved.
@end deftypecv

@deftypecv {output member} gdbm_recovery size_t failed_buckets
Number of buckets that could not be retrieved.
@end deftypecv

@deftypecv {output member} gdbm_recovery {char *} backup_name
@kwindex GDBM_RCVR_BACKUP
Name of the file keeping the copy of the original database, in the
state prior to recovery.  It is filled if the @var{GDBM_RCVR_BACKUP}
flag is set.  The string is allocated using the @code{malloc} call.
The caller is responsible for freeing that memory when no longer needed.
@end deftypecv
@end deftypefn

@kwindex GDBM_RCVR_FORCE
By default, @code{gdbm_recovery} first checks the database for
inconsistencies and attempts recovery only if some were found.
The special flag bit @code{GDBM_RCVR_FORCE} instructs
@code{gdbm_recovery} to omit this check and to perform database recovery
unconditionally.

@node Crash Tolerance
@chapter Crash Tolerance

Crash tolerance is a new (as of release 1.21) feature that can be
enabled at compile time, and used in environments with appropriate
support from the OS and the filesystem.  As of version
@value{VERSION}, this means a Linux kernel 5.12.12 or later and
a filesystem that supports reflink copying, such as XFS, BtrFS, or
OCFS2.  If these prerequisites are met, crash tolerance code will
be enabled automatically by the @command{configure} script when
building the package.

The crash-tolerance mechanism, when used correctly, guarantees that a
logically consistent (@pxref{Database consistency}) recent state of
application data can be recovered following a crash.  Specifically, it
guarantees that the state of the database file corresponding to the
most recent successful @code{gdbm_sync} call can be recovered.

If the new mechanism is used correctly, crashes such as power
outages, OS kernel panics, and (some) application process crashes
will be tolerated.  Non-tolerated failures include physical
destruction of storage devices and corruption due to bugs in
application logic.  For example, the new mechanism won't help if a
pointer bug in your application corrupts @command{GDBM}'s private in-memory
data which in turn corrupts the database file.

In the following sections we will describe how to enable crash
tolerance in your application and what to do if a crash occurs.

The design rationale of the crash tolerance mechanism is described in
detail in the article, @cite{Crashproofing the Original NoSQL Key-Value
Store}, by Terence Kelly, @cite{ACM Queue magazine}, July/August 2021,
available from the @uref{https://queue.acm.org/DrillBits5/, ACM Digital Library}.
If you have difficulty retrieving this paper, please contact the
author at @email{tpkelly@@acm.org}, @email{tpkelly@@cs.princeton.edu},
or @email{tpkelly@@eecs.umich.edu}.

@node Filesystems supporting crash tolerance
@section Using Proper Filesystem

Use a filesystem that supports reflink copying.  Currently XFS, BtrFS,
and OCFS2 support reflink.  You can create such a filesystem if you
don't have one already.  (Note that reflink support may require that
special options be specified at the time of filesystem creation; this
is true of XFS.)  The most conventional way to create a filesystem is
on a dedicated storage device.  However it is also possible to create
a filesystem @emph{within an ordinary file} on some other filesystem.

For example, the following commands, executed as root, will create a
smallish XFS filesystem inside a file on another filesystem:

@example
mkdir XFS
cd XFS
truncate --size 512m XFSfile
mkfs -t xfs -m crc=1 -m reflink=1 XFSfile
mkdir XFSmountpoint
mount -o loop XFSfile XFSmountpoint
@end example

The XFS filesystem is now available in directory
@file{XFSmountpoint}.  Now, create a directory where your
unprivileged user account may create and delete files:

@example
cd XFSmountpoint
mkdir test
chown @var{user}:@var{group} test
@end example

@noindent
(where @var{user} and @var{group} are the user and group names of the
unprivileged account the application uses).

Reflink copying via @code{ioctl(FICLONE)} should work for files in and
below this directory.  You can test reflink copying using the GNU
@command{cp} program:

@example
cp --reflink=always file1 file2
@end example

@xref{cp invocation, reflink, reflink, coreutils, @sc{gnu} Coreutils}.

Your GNU dbm database file and two @dfn{snapshot} files described below must
all reside on the same reflink-capable filesystem.

@node Enabling crash tolerance
@section Enabling crash tolerance

Open a GNU dbm database with @code{gdbm_open}.  Whenever possible, use
the extended @command{GDBM} format (@pxref{Numsync}).  Generally
speaking, this means using the @code{GDBM_NUMSYNC} flag when creating
the database.  Unless you know what you are doing, do not specify
the @code{GDBM_SYNC} flag when opening the database.  The reason is that
you want your application to explicitly control when @code{gdbm_sync}
is called; you don't want an implicit sync on every database
operation (@pxref{Sync}).

Request crash tolerance by invoking the following interface:

@example
int gdbm_failure_atomic (GDBM_FILE @var{dbf}, const char *@var{even},
                         const char *@var{odd});
@end example

The @var{even} and @var{odd} arguments are the pathnames of two files that
will be created and filled with @dfn{snapshots} of the database file.
These two files must not exist when @code{gdbm_failure_atomic} is
called and must reside on the same reflink-capable filesystem as the
database file.

After you call @code{gdbm_failure_atomic}, every call to
@code{gdbm_sync} will make an efficient reflink snapshot of the
database file in either the @var{even} or the @var{odd} snapshot file;
consecutive @code{gdbm_sync} calls alternate between the two, hence
the names.  The permission bits and @code{mtime} timestamps on the
snapshot files determine which one contains the state of the database
file corresponding to the most recent successful @code{gdbm_sync}.
@xref{Crash recovery}, for discussion of crash recovery.

@node Synchronizing the Database
@section Synchronizing the Database

When your application knows that the state of the database is
consistent (i.e., all relevant application-level invariants hold),
you may call @code{gdbm_sync}.  For example, if your application
manages bank accounts, transferring money from one account to another
should maintain the invariant that the sum of the two accounts is the
same before and after the transfer: It is correct to decrement account
@samp{A} by $7, increment account @samp{B} by $7, and then call
@code{gdbm_sync}.  However it is @emph{not} correct to call
@code{gdbm_sync} @emph{between} the decrement of @samp{A} and the
increment of @samp{B}, because a crash immediately after that call
would destroy money.  The general rule is simple, sensible, and
memorable: Call @code{gdbm_sync} only when the database is in a state
from which you are willing and able to recover following a crash.  (If
you think about it you'll realize that there's never any other moment
when you'd really want to call @code{gdbm_sync}, regardless of whether
crash-tolerance is enabled.  Why on earth would you push the state of
an inconsistent unrecoverable database down to durable media?).

@node Crash recovery
@section Crash recovery

If a crash occurs, the snapshot file (@var{even} or @var{odd})
containing the database state reflecting the most recent successful
@code{gdbm_sync} call is the snapshot file whose permission bits are
read-only and whose last-modification timestamp is greatest.  If both
snapshot files are readable, we choose the one with the most recent
last-modification timestamp.  Modern operating systems record
timestamps in nanoseconds, which gives sufficient confidence that the
timestamps of the two snapshots will differ.  However, one can't rule
out the possibility that the two snapshot files will both be readable
and have identical timestamps@footnote{This can happen, for example,
if the storage is very fast and the system clock is low-resolution, or
if the system administrator sets the system clock backwards.  In the
latter case one can end up with the most recent snapshot file having
modification time earlier than that of the obsolete snapshot.}.  To
cope with this, @command{GDBM} version 1.21 introduced the new
@dfn{extended database format}, which stores in the database file
header the number of synchronizations performed so far.  This number
can reliably be used to select the most recent snapshot, independently
of its timestamp.  We strongly suggest using this new format when
writing crash-tolerant applications.  @xref{Numsync}, for a detailed
discussion.

The @code{gdbm_latest_snapshot} function is provided, that selects the
right snapshot among the two.  Invoke it as:

@example
@group
const char *recovery_file = NULL;
result = gdbm_latest_snapshot (even, odd, &recovery_file);
@end group
@end example

@noindent
where @var{even} and @var{odd} are names of the snapshot files.  On
success, it stores the pointer to the most recent snapshot file name
in @var{recovery_file} and returns @code{GDBM_SNAPSHOT_OK}.  To
finalize the recovery, rename this file to the name of your database
file and re-open it using @code{gdbm_open}.  You should discard the
remaining snapshot.

If an error occurs, @code{gdbm_latest_snapshot} returns one of the
following error codes.

@defvr {gdbm_latest_snapshot} GDBM_SNAPSHOT_BAD
Neither snapshot file is readable.  This means that the crash has occurred
before @code{gdbm_failure_atomic} completed.  In this case, it is best
to fall back on a safe backup copy of the data file.
@end defvr

@defvr {gdbm_latest_snapshot} GDBM_SNAPSHOT_ERR
System error occurred in @code{gdbm_latest_snapshot}.  Examine the
system @code{errno} variable for details.  Its possible values are:

@table @code
@item EACCES
The file mode of one of the snapshot files was incorrect.  Each snapshot
file can be either readable (0400) or writable (0200), but not both.
This probably means that someone touched one or both snapshot files
after the crash and before your attempt to recover from it.  This case
needs additional investigation.  If you're sure that the only change
someone made to the files is altering their modes, and your database
is in @dfn{numsync} format (@pxref{Numsync}), you can reset the modes
to 0400 and retry the recovery.

This error can also be returned by underlying @code{stat} call,
meaning that search permission was denied for one of the directories
in the path prefix of a snapshot file name.  That again means that
someone has messed with permissions after the crash.

@item EINVAL
Some arguments passed to @code{gdbm_latest_snapshot} were not valid.
It is a programmer's error which means that your application needs to be
fixed.

@item ENOSYS
Function is not implemented.  This means @command{GDBM} was built without
crash-tolerance support.

@item Other value (@code{EBADF}, @code{EFAULT}, etc)
An error occurred when trying to @code{stat} the snapshot file.
@xref{ERRORS,,,stat(2),stat(2) man page}, for a discussion of
possible @code{errno} values.
@end table
@end defvr

@defvr {gdbm_latest_snapshot} GDBM_SNAPSHOT_SAME
File modes and modification dates of both snapshot files are exactly
the same.  This can happen only if numsync is not available
(@pxref{Numsync}).
@end defvr

@defvr {gdbm_latest_snapshot} GDBM_SNAPSHOT_SUSPICIOUS
For the database in extended @dfn{numsync} format (@pxref{Numsync}):
the @code{numsync} values of the two snapshot differ by more than
one.  Check the arguments to the @code{gdbm_latest_snapshot} function.
The most probably reason of such an error is that the @var{even} and
@var{odd} parameters point to snapshot files belonging to different
database files.
@end defvr

If you get any of these errors, we strongly suggest to undertake
@dfn{manual recovery}.

@node Manual crash recovery
@section Manual crash recovery

@dfn{Manual recovery} is usually performed with the help of the
@command{gdbmtool} utility.  Start @command{gdbmtool} in read-only
mode (the @option{-r}) option.  Once in the command shell, issue the
following command:

@example
snapshot @var{a} @var{b}
@end example

@noindent
where @var{a} and @var{b} are names of the two snapshot files you
configured using the @code{gdbm_failure_atomic} function.  This
command investigates both files and prints out detailed
diagnostics.

Its output begins with a line listing one of the error codes above,
followed by a colon and a textual description of the error.  The lines
that follow show details for each snapshot file.

Each snapshot description begins with the snapshot file name followed
by a colon and four fields, in this order:

@enumerate 1
@item File permission bits in octal.
@item File permission bits in @command{ls -l} notation.
@item Modification timestamp.
@item Numsync counter.
For databases in standard @command{GDBM} format, this field is @samp{N/A}.  If
the counter cannot be obtained because of error, this field is @samp{?}.
@end enumerate

Any errors or inconsistencies discovered are reported in the lines
that follow, one error per line.  Here's an example of the
@command{snapshot} command output, describing the
@code{GDBM_SNAPSHOT_ERR} condition:

@example
@group
gdbmtool> snapshot even.dbf odd.dbf
GDBM_SNAPSHOT_ERR: Error selecting snapshot.
even.dbf: 200 -w------- 1627820627.485681330 ?
odd.dbf: 600 rw------- 1627820627.689503918 301
odd.dbf: ERROR: bad file mode
@end group
@end example

Line 2 lists the meta-data of the snapshot @file{even.dbf}.  The
@code{numsync} field contains question mark because the file
permissions (write-only) prevented @command{gdbmtool} from opening it.

The lines for @file{odd.dbf} show the actual reason for the error: bad
file mode (read-write).  Apparently, the file mode has been changed
manually after the crash.  The timestamp of the file, which is more
recent than that of @file{even.dbf}, suggests that it might be used for
recovery.  To confirm this guess, change the mode of the
@file{even.dbf} to read-only and repeat the @command{snapshot} command:

@example
@group
gdbmtool> ! chmod 400 even.dbf
gdbmtool> snapshot even.dbf odd.dbf
GDBM_SNAPSHOT_ERR: Error selecting snapshot.
even.dbf: 400 r-------- 1627820627.485681330 300
odd.dbf: 600 rw------- 1627820627.689503918 301
odd.dbf: ERROR: bad file mode
@end group
@end example

This shows the numsync value of the @file{even.dbf} file, which is
exactly one less than that of @file{odd.dbf}.  This means that the
latter should be selected for recovery.

For completeness sake, you can change the mode of @file{odd.dbf} to
read-only as well and repeat the @command{snapshot} command.  In this
case you will see:

@example
@group
gdbmtool> ! chmod 400 odd.dbf
gdbmtool> snapshot even.dbf odd.dbf
GDBM_SNAPSHOT_OK: Selected the most recent snapshot.
odd.dbf: 400 r-------- 1627820627.689503918 301
@end group
@end example

@node Performance Impact
@section Performance Impact

The purpose of a parachute is not to hasten descent.  Crash tolerance
is a safety mechanism, not a performance accelerator.  Reflink
copying is designed to be as efficient as possible, but making
snapshots of the GNU dbm database file on every @code{gdbm_sync} call
entails overheads.  The performance impact of @command{GDBM} crash tolerance
will depend on many factors including the type and configuration of
the underlying storage system, how often the application calls
@code{gdbm_sync}, and the extent of changes to the database file
between consecutive calls to @code{gdbm_sync}.

@node Availability
@section Availability

To ensure that application data can survive the failure of one or
more storage devices, replicated storage (e.g., RAID) may be used
beneath the reflink-capable filesystem.  Some cloud providers offer
block storage services that mimic the interface of individual storage
devices but that are implemented as high-availability fault-tolerant
replicated distributed storage systems.  Installing a reflink-capable
filesystem atop a high-availability storage system is a good starting
point for a high-availability crash-tolerant @command{GDBM}.

@node Numsync
@section Numsync Extension

In @ref{Crash recovery}, we have shown that for database recovery,
one should select the snapshot whose permission bits are read-only and
whose last-modification timestamp is greatest.  However, there may be
cases when a crash occurs at such a time that both snapshot files
remain readable.  It may also happen, that their permissions had
been reset to read-only and/or modification times inadvertently
changed before recovery.  To make it possible to select the right
snapshot in such cases, a new @dfn{extended database format} was
introduced in @command{GDBM} version 1.21.  This format adds to the
database header the @code{numsync} field, which holds the number of
synchronizations the database underwent before being closed or
abandoned due to a crash.

A readable snapshot is a consistent copy of the database at a given point of
time.  Thus, if both snapshots of a database in extended format are
readable, it will suffice to examine their @code{numsync} counters
and select the one whose @code{numsync} is greater.  That's what
the @code{gdbm_latest_snapshot} function does in this case.

It is worth noticing, that the two counters should differ exactly by
one.  If the difference is greater than that, @code{gdbm_latest_snapshot}
will return a special status code, @code{GDBM_SNAPSHOT_SUSPICIOUS}.
If, during a recovery attempt, you get this status code, we recommend
to proceed with the manual recovery (@pxref{Manual crash recovery}).

To create a database in extended format, call @code{gdbm_open} with
both @code{GDBM_NEWDB} and @code{GDBM_NUMSYNC} flags:

@example
dbf = gdbm_open(dbfile, 0, GDBM_NEWDB|GDBM_NUMSYNC, 0600, NULL);
@end example

@noindent
Notice, that this flag must always be used together with
@code{GDBM_NEWDB} (@pxref{Open}).  It is silently ignored when used
together with another opening flag.

A standard @command{GDBM} database can be converted to the extended
format and vice versa.  To convert an existing database to the
extended format, use the @code{gdbm_convert} function (@pxref{Database
format}):

@example
  rc = gdbm_convert(dbf, GDBM_NUMSYNC);
@end example

You can do the same using the @command{gdbmtool} utility
(@pxref{commands, upgrade}):

@example
gdbmtool @var{dbname} upgrade
@end example

To convert a database from extended format back to the standard
@command{GDBM} format, do:

@example
  rc = gdbm_convert(dbf, 0);
@end example

To do the same from the command line, run:

@example
gdbmtool @var{dbname} downgrade
@end example

@node Crash Tolerance API
@section Crash Tolerance API

@deftypefn {gdbm interface} int gdbm_failure_atomic (GDBM_FILE @var{dbf}, @
   const char *@var{even}, const char *@var{odd})
Enables crash tolerance for the database file @var{dbf}.  The
@var{even} and @var{odd} arguments are the pathnames of two files that
will be created and filled with snapshots of the database file.
These two files must not exist when @code{gdbm_failure_atomic} is
called and must reside on the same reflink-capable filesystem as the
database file.

Returns 0 on success.  On failure, returns -1 and sets
@code{gdbm_errno} to one of the following values:

@table @code
@item GDBM_ERR_USAGE
Improper function usage.  Either @var{even} or @var{odd} is
@code{NULL}, or they point to the same string.

@item GDBM_NEED_RECOVERY
The database needs recovery.  @xref{Recovery}.

@item GDBM_ERR_SNAPSHOT_CLONE
Failed to clone the database file into a snapshot.  Examine the system
@code{errno} variable for details.
@end table

If one of the following error codes is returned, examine the system
@code{errno} variable for details:

@table @code
@item GDBM_ERR_REALPATH
Call to @code{realpath} function failed.  @code{realpath} is used to
determine actual path names of the snapshot files.

@item GDBM_FILE_OPEN_ERROR
Unable to create snapshot file.

@item GDBM_FILE_SYNC_ERROR
Failed to sync a snapshot file or one of directories in its pathname,
during initial synchronization.

@item GDBM_FILE_CLOSE_ERROR
Failed to close a snapshot file or one of directories in its pathname,
during initial synchronization.

@item GDBM_ERR_FILE_MODE
The @code{fchmod} call on one of the snapshot files failed.
@end table

Notes:

@itemize @bullet
@item It is not an error to call @code{gdbm_failure_atomic} several times.
Each subsequent call closes the previously configured snapshot files
and installs new ones instead.

@item Crash tolerance settings are cleared by functions
@code{gdbm_recover} (@pxref{Recovery}) and @code{gdbm_reorganize}
(@pxref{Reorganization}).  In case of @code{gdbm_recover}, it should
not be a problem, because if you enabled crash tolerance, the
procedure described in @ref{Crash recovery} is the preferred way of
recovering the database.  If, however, you decided to call either
function even though you had enabled crash tolerance previously, be
sure to call @code{gdbm_failure_atomic} again with the same arguments
as before (provided that the call returns successfully).
@end itemize
@end deftypefn

@deftypefn {gdbm interface} int gdbm_latest_snapshot (const char *@var{even}, @
  const char *@var{odd}, const char **@var{retval})
@kwindex GDBM_SNAPSHOT_OK
@kwindex GDBM_SNAPSHOT_BAD
@kwindex GDBM_SNAPSHOT_ERR
@kwindex GDBM_SNAPSHOT_SAME
Selects between two snapshots, @var{even} and @var{odd}, the one to be
used for crash recovery.  On success, stores a pointer to the selected
filename in the memory location pointed to by @var{retval} and returns
@code{GDBM_SNAPSHOT_OK}.  If neither snapshot file is usable, the
function returns @code{GDBM_SNAPSHOT_BAD}.  If a system error occurs, it
returns @code{GDBM_SNAPSHOT_ERR} and sets @code{errno} to the error code
describing the problem.  Finally, in the unlikely case that it cannot
select between the two snapshots (this means they are both readable
and have exactly the same @code{mtime} timestamp), the function returns
@code{GDBM_SNAPSHOT_SAME}.

@kwindex GDBM_SNAPSHOT_SUSPICIOUS
If the @samp{numsync} extension is enabled (@pxref{Numsync}), the
function can also return the @code{GDBM_SNAPSHOT_SUSPICIOUS} status
code.  This happens when the @code{numsync} counters in the two
snapshots differ by more than one.

@xref{Crash recovery}, for a detailed description of possible return
codes and their interpretation.

If any value other than @code{GDBM_SNAPSHOT_OK} is returned, it is
guaranteed that the function did not touch @var{retval}.    In this case
it is recommended to switch to manual recovery procedure, letting the
user examine the snapshots and take the appropriate action.
@pxref{Manual crash recovery}, for details.
@end deftypefn

@node Options
@chapter Setting options
@cindex database options
@cindex options, database

@command{GDBM} supports the ability to set certain options on an already
open database.

@deftypefn {gdbm interface} int gdbm_setopt (GDBM_FILE @var{dbf}, int @var{option}, @
           void *@var{value}, int @var{size})
Sets an option on the database or returns the value of an option.

The parameters are:

@table @var
@item dbf
The pointer returned by @code{gdbm_open}.
@item option
The option to be set or retrieved.
@item value
A pointer to the value to which @var{option} will be set or where to
place the option value (depending on the option).
@item size
The length of the data pointed to by @var{value}.
@end table

The return value will be @code{-1} upon failure, or @code{0} upon
success.  The global variable @code{gdbm_errno} will be set upon failure.
@end deftypefn

The valid options are:

@defvr {Option} GDBM_SETCACHESIZE
@defvrx {Option} GDBM_CACHESIZE
Set the size of the internal bucket cache.  The @var{value} should
point to a @code{size_t} holding the desired cache size, or the
constant @code{GDBM_CACHE_AUTO}, to adjust the cache size
automatically.

By default, a newly open database is configured to dynamically
accommodate the cache size to the number of index buckets in the
database file.  This provides for the best performance.

If another @var{value} is set, it is adjusted to the nearest larger
power of two.

Use this option if you wish to limit the memory usage at the expense
of performance.  If you chose to do so, please bear in mind that cache
becomes effective when its size is greater then 2/3 of the number of
index bucket counts in the database.  The best performance results are
achieved when cache size equals the number of buckets.  For example:

@example
size_t bn;
gdbm_bucket_count (dbf, &bn);
ret = gdbm_setopt (dbf, GDBM_SETCACHESIZE, &bn, sizeof (bn));
@end example

To request the automatically adjustable cache size, use the constant
@code{GDBM_CACHE_AUTO}:

@example
size_t bn = GDBM_CACHE_AUTO;
ret = gdbm_setopt (dbf, GDBM_SETCACHESIZE, &bn, sizeof (bn));
@end example
@end defvr

@defvr {Option} GDBM_GETCACHESIZE
Return the actual size of the internal bucket cache.  The @var{value}
should point to a @code{size_t} variable, where the size will be
stored.
@end defvr

@defvr {Option} GDBM_SETCACHEAUTO
Controls whether the cache size will be adjusted automatically as
needed.  The @var{value} should point to an integer: @code{TRUE} to
enable automatic cache adjustment and @code{FALSE} to disable it.

The following two calls are equivalent:

@example
int t = TRUE;
gdbm_setopt (dbf, GDBM_SETCACHEAUTO, &t, sizeof (t));

size_t n = GDBM_CACHE_AUTO;
gdbm_setopt (dbf, GDBM_SETCACHESIZE, &n, sizeof (n));
@end example
@end defvr

@defvr {Option} GDBM_GETCACHEAUTO
Return the state of the automatic cache size adjustment.  The
@var{value} should point to an integer which, upon successful return,
will have the value @code{TRUE} if the automatic cache size adjustment
is enabled and @code{FALSE} otherwise.
@end defvr

@defvr {Option} GDBM_GETFLAGS
Return the flags describing the state of the database.  The @var{value} should
point to an @code{int} variable where to store the flags.  On success,
its value will be similar to the flags used when opening the database
(@pxref{Open, gdbm_open}), except that it will reflect the current state
(which may have been altered by another calls to @code{gdbm_setopt}).
@end defvr

@defvr {Option} GDBM_GETDBFORMAT
Return the database format.  The @var{value} should point to an
@code{int} variable.  Upon successful return, it will be set to
@samp{0} if the database is in standard format and @code{GDBM_NUMSYNC}
if it is in extended format.  @xref{Database format}.
@end defvr

@defvr {Option} GDBM_GETDIRDEPTH
Returns the @dfn{directory depth}: the number of initial (most significant)
bits in hash value that are interpreted as index to the directory.  The
actual directory size can be computed as @code{1 << @var{value}}.

The @var{value} argument should point to an @code{int}.  
@end defvr

@defvr {Option} GDBM_GETBUCKETSIZE
Returns the @dfn{bucket capacity}: maximum number of keys per bucket
(@code{int}).
@end defvr

@defvr {Option} GDBM_FASTMODE
Enable or disable the @dfn{fast writes mode}, i.e.@: writes without
subsequent synchronization.  The @var{value} should point
to an integer: @code{TRUE} to enable fast mode, and @code{FALSE} to
disable it.

This option is retained for compatibility with previous versions of
@command{GDBM}.  Its effect is the reverse of @code{GDBM_SETSYNCMODE}.
@end defvr

@defvr {Option} GDBM_SETSYNCMODE
@defvrx {Option} GDBM_SYNCMODE
Turn on or off file system synchronization operations.  This
setting defaults to off.  The @var{value} should point
to an integer: @code{TRUE} to turn synchronization on, and @code{FALSE} to
turn it off.

Note, that this option is a reverse of @code{GDBM_FASTMODE},
i.e.@: calling @code{GDBM_SETSYNCMODE} with @code{TRUE} has the same effect
as calling @code{GDBM_FASTMODE} with @code{FALSE}.

The @code{GDBM_SYNCMODE} option is provided for compatibility with
earlier versions.
@end defvr

@defvr {Option} GDBM_GETSYNCMODE
Return the current synchronization status.  The @var{value} should
point to an @code{int} where the status will be stored.
@end defvr

@defvr {Option} GDBM_SETCENTFREE
@defvrx {Option} GDBM_CENTFREE
@emph{NOTICE: This feature is still under study.}

Set central free block pool to either on or off.  The default is off,
which is how previous versions of @command{GDBM} handled free blocks.  If
set, this option causes all subsequent free blocks to be placed in the
@emph{global} pool, allowing (in theory) more file space to be reused
more quickly.  The @var{value} should point to an integer: @code{TRUE} to
turn central block pool on, and @code{FALSE} to turn it off.

The @code{GDBM_CENTFREE} option is provided for compatibility with
earlier versions.
@end defvr

@defvr {Option} GDBM_SETCOALESCEBLKS
@defvrx {Option} GDBM_COALESCEBLKS
@emph{NOTICE: This feature is still under study.}

Set free block merging to either on or off.  The default is off, which
is how previous versions of @command{GDBM} handled free blocks.  If set,
this option causes adjacent free blocks to be merged.  This can become
a @acronym{CPU} expensive process with time, though, especially if
used in conjunction with GDBM_CENTFREE.  The @var{value} should point
to an integer: @code{TRUE} to turn free block merging on, and @code{FALSE} to
turn it off.
@end defvr

@defvr {Option} GDBM_GETCOALESCEBLKS
Return the current status of free block merging.  The @var{value} should
point to an @code{int} where the status will be stored.
@end defvr

@defvr {Option} GDBM_SETMAXMAPSIZE
Sets maximum size of a memory mapped region.  The @var{value} should
point to a value of type @code{size_t}, @code{unsigned long} or
@code{unsigned}.  The actual value is rounded to the nearest page
boundary (the page size is obtained from
@code{sysconf(_SC_PAGESIZE)}).
@end defvr

@defvr {Option} GDBM_GETMAXMAPSIZE
Return the maximum size of a memory mapped region.  The @var{value} should
point to a value of type @code{size_t} where to return the data.
@end defvr

@defvr {Option} GDBM_SETMMAP
Enable or disable memory mapping mode.  The @var{value} should point
to an integer: @code{TRUE} to enable memory mapping or @code{FALSE} to
disable it.
@end defvr

@defvr {Option} GDBM_GETMMAP
Check whether memory mapping is enabled.  The @var{value} should point
to an integer where to return the status.
@end defvr

@defvr {Option} GDBM_GETDBNAME
Return the name of the database disk file.  The @var{value} should
point to a variable of type @code{char**}.  A pointer to the newly
allocated copy of the file name will be placed there.  The caller is
responsible for freeing this memory when no longer needed.  For
example:

@example
char *name;

if (gdbm_setopt (dbf, GDBM_GETDBNAME, &name, sizeof (name)))
  @{
     fprintf (stderr, "gdbm_setopt failed: %s\n",
              gdbm_strerror (gdbm_errno));
  @}
else
  @{
    printf ("database name: %s\n", name);
    free (name);
  @}
@end example
@end defvr

@defvr {Option} GDBM_GETBLOCKSIZE
Return the block size in bytes.  The @var{value} should point to @code{int}.
@end defvr

@node Locking
@chapter File Locking
@cindex locking

@kwindex GDBM_NOLOCK
With locking disabled (if @code{gdbm_open} was called with @code{GDBM_NOLOCK}),
the user may want to perform their own file locking on the database file
in order to prevent multiple writers operating on the same file
simultaneously.

In order to support this, the @code{gdbm_fdesc} routine is provided.

@deftypefn {gdbm interface} int gdbm_fdesc (GDBM_FILE @var{dbf})
Returns the file descriptor of the database @var{dbf}.  This value
can be used as an argument to @code{flock}, @code{lockf} or similar
calls.
@end deftypefn

@node Variables
@chapter Useful global variables

The following global variables and constants are available:

@deftypevar gdbm_error gdbm_errno
This variable contains error code from the last failed @command{GDBM}
call.  @xref{Error codes}, for a list of available error codes and
their descriptions.

Use @code{gdbm_strerror} (@pxref{Errors}) to convert it to a
descriptive text.
@end deftypevar

@deftypevar {const char *} gdbm_errlist[]
This variable is an array of error descriptions, which is used by
@code{gdbm_strerror} to convert error codes to human-readable text
(@pxref{Errors}).  You can access it directly, if you wish so.  It
contains @code{_GDBM_MAX_ERRNO + 1} elements and can be directly
indexed by the error code to obtain a corresponding descriptive
text.
@end deftypevar

@deftypevar {int const} gdbm_syserr[]
Array of boolean values indicating, for each @command{GDBM} error code, whether
the value of @code{errno}(3) variable is meaningful for this error
code.  @xref{gdbm_check_syserr}.
@end deftypevar

@defvr {Constant} _GDBM_MIN_ERRNO
The minimum error code used by @command{GDBM}.
@end defvr

@defvr {Constant} _GDBM_MAX_ERRNO
The maximum error code used by @command{GDBM}.
@end defvr

@cindex version number
@deftypevar {const char *} gdbm_version
A string containing the version information.
@end deftypevar

@deftypevar {int const} gdbm_version_number[3]
This variable contains the @command{GDBM} version numbers:

@multitable @columnfractions 0.4 0.5
@headitem Index @tab Meaning
@item 0         @tab Major number
@item 1         @tab Minor number
@item 2         @tab Patchlevel number
@end multitable

Additionally, the following constants are defined in the @file{gdbm.h}
file:

@table @asis
@kwindex GDBM_VERSION_MAJOR
@item GDBM_VERSION_MAJOR
Major number.

@kwindex GDBM_VERSION_MINOR
@item GDBM_VERSION_MINOR
Minor number.

@kwindex GDBM_VERSION_PATCH
@item GDBM_VERSION_PATCH
Patchlevel number.
@end table

These can be used to verify whether the header file matches the library.
@end deftypevar

To compare two split-out version numbers, use the following function:

@deftypefn {gdbm interface} int gdbm_version_cmp (int const @var{a}[3], @
                                                  int const @var{b}[3])
Compare two version numbers.  Return @code{-1} if @var{a} is less than
@var{b}, @code{1} if @var{a} is greater than @var{b} and @code{0} if
they are equal.

Comparison is done from left to right, so that:

@example
a = @{ 1, 8, 3 @};
b = @{ 1, 8, 3 @};
gdbm_version_cmp (a, b) @result{} 0

a = @{ 1, 8, 3 @};
b = @{ 1, 8, 2 @};
gdbm_version_cmp (a, b) @result{} 1

a = @{ 1, 8, 3 @};
b = @{ 1, 9. 0 @};
gdbm_version_cmp (a, b) @result{} -1
@end example
@end deftypefn

@node Additional functions
@chapter Additional functions

@deftypefn {gdbm interface} int gdbm_avail_verify (GDBM_FILE @var{dbf})
Verify if the available block stack is in consistent state.  On
success, returns 0.  If any errors are encountered, sets the
@code{gdbm_errno} to @code{GDBM_BAD_AVAIL}, marks the database as
needing recovery (@pxref{Recovery}) and return -1.
@end deftypefn

@node Error codes
@chapter Error codes
@cindex error codes

This chapter summarizes error codes which can be set by the
functions in @command{GDBM} library.

@defvr {Error Code} GDBM_NO_ERROR
No error occurred.
@end defvr

@defvr {Error Code} GDBM_MALLOC_ERROR
Memory allocation failed.  Not enough memory.
@end defvr

@defvr {Error Code} GDBM_BLOCK_SIZE_ERROR
@kwindex GDBM_BSEXACT
This error is set by the @code{gdbm_open} function (@pxref{Open}), if
the value of its @var{block_size} argument is incorrect and the
@code{GDBM_BSEXACT} flag is set.
@end defvr

@defvr {Error Code} GDBM_FILE_OPEN_ERROR
The library was not able to open a disk file.  This can be set by
@code{gdbm_open} (@pxref{Open}), @code{gdbm_dump} (@code{gdbm_export}) and
@code{gdbm_load} (@code{gdbm_import}) functions (@pxref{Flat files}).

Inspect the value of the system @code{errno} variable to get more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_FILE_WRITE_ERROR
Writing to a disk file failed.  This can be set by
@code{gdbm_open} (@pxref{Open}), @code{gdbm_dump} (@code{gdbm_export}) and
@code{gdbm_load} (@code{gdbm_import}) functions.

Inspect the value of the system @code{errno} variable to get more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_FILE_SEEK_ERROR
Positioning in a disk file failed.  This can be set by
@code{gdbm_open} (@pxref{Open}) function.

Inspect the value of the system @code{errno} variable to get a more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_FILE_READ_ERROR
Reading from a disk file failed.  This can be set by
@code{gdbm_open} (@pxref{Open}), @code{gdbm_dump} (@code{gdbm_export}) and
@code{gdbm_load} (@code{gdbm_import}) functions.

Inspect the value of the system @code{errno} variable to get a more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_BAD_MAGIC_NUMBER
The file given as argument to @code{gdbm_open} function is not a valid
@command{GDBM} file: it has a wrong magic number.
@end defvr

@defvr {Error Code} GDBM_EMPTY_DATABASE
The file given as argument to @code{gdbm_open} function is not a valid
@command{GDBM} file: it has zero length.
@end defvr

@defvr {Error Code} GDBM_CANT_BE_READER
This error code is set by the @code{gdbm_open} function if it is not
able to lock file when called in @code{GDBM_READER} mode (@pxref{Open,
GDBM_READER}).
@end defvr

@defvr {Error Code} GDBM_CANT_BE_WRITER
This error code is set by the @code{gdbm_open} function if it is not
able to lock file when called in writer mode (@pxref{Open}).
@end defvr

@defvr {Error Code} GDBM_READER_CANT_DELETE
Set by the @code{gdbm_delete} (@pxref{Delete}) if it attempted to
operate on a database that is open in read-only mode (@pxref{Open,
GDBM_READER}).
@end defvr

@defvr {Error Code} GDBM_READER_CANT_STORE
Set by the @code{gdbm_store} (@pxref{Store}) if it attempted to
operate on a database that is open in read-only mode (@pxref{Open,
GDBM_READER}).
@end defvr

@defvr {Error Code} GDBM_READER_CANT_REORGANIZE
Set by the @code{gdbm_reorganize} (@pxref{Reorganization}) if it attempted to
operate on a database that is open in read-only mode (@pxref{Open,
GDBM_READER}).
@end defvr

@defvr {Error Code} GDBM_ITEM_NOT_FOUND
Requested item was not found.  This error is set by @code{gdbm_delete}
(@pxref{Delete}) and @code{gdbm_fetch} (@pxref{Fetch}) when the requested
@var{key} value is not found in the database.
@end defvr

@defvr {Error Code} GDBM_REORGANIZE_FAILED
The @code{gdbm_reorganize} function is not
able to create a temporary database.  @xref{Reorganization}.
@end defvr

@defvr {Error Code} GDBM_CANNOT_REPLACE
Cannot replace existing item.  This error is set by the
@code{gdbm_store} if the requested @var{key} value is found in the
database and the @var{flag} parameter is not @code{GDBM_REPLACE}.
@xref{Store}, for a detailed discussion.
@end defvr

@defvr {Error Code} GDBM_MALFORMED_DATA
@defvrx {Error Code} GDBM_ILLEGAL_DATA
Input data was malformed in some way.  When returned by
@code{gdbm_load}, this means that the input file was not a valid
@command{GDBM} dump file (@pxref{gdbm_load function}).  When returned by
@code{gdbm_store}, this means that either @var{key} or @var{content}
parameter had its @code{dptr} field set to @code{NULL}
(@pxref{Store}).

The @code{GDBM_ILLEGAL_DATA} is an alias for this error code,
maintained for backward compatibility.  Its use in modern applications
is discouraged.
@end defvr

@defvr {Error Code} GDBM_OPT_ALREADY_SET
Requested option can be set only once and was already set.  As of
version @value{VERSION}, this error code is no longer used.  In prior
versions it could have been returned by the @code{gdbm_setopt}
function when setting the @code{GDBM_CACHESIZE} value.
@end defvr

@defvr {Error Code} GDBM_OPT_BADVAL
@defvrx {Error Code} GDBM_OPT_ILLEGAL
The @var{option} argument is not valid or the @var{value} argument
points to an invalid value in a call to @code{gdbm_setopt} function.
@xref{Options}.

@code{GDBM_OPT_ILLEGAL} is an alias for this error code, maintained
for backward compatibility.  Modern applications should not use it.
@end defvr

@defvr {Error Code} GDBM_BYTE_SWAPPED
The @code{gdbm_open} function (@pxref{Open}) attempts to open a
database which is created on a machine with different byte ordering.
@end defvr

@defvr {Error Code} GDBM_BAD_FILE_OFFSET
The @code{gdbm_open} function (@pxref{Open}) sets this error code if
the file it tries to open has a wrong magic number.
@end defvr

@defvr {Error Code} GDBM_BAD_OPEN_FLAGS
Set by the @code{gdbm_dump} (@code{gdbm_export}) function if supplied
an invalid @var{flags} argument.  @xref{Flat files}.
@end defvr

@defvr {Error Code} GDBM_FILE_STAT_ERROR
Getting information about a disk file failed.  The system @code{errno}
will give more details about the error.

This error can be set by the following functions: @code{gdbm_open},
@code{gdbm_reorganize}.
@end defvr

@defvr {Error Code} GDBM_FILE_EOF
End of file was encountered where more data was expected to be
present.  This error can occur when fetching data from the database
and usually means that the database is truncated or otherwise corrupted.

This error can be set by any @command{GDBM} function that does I/O.  Some of
these functions are: @code{gdbm_delete}, @code{gdbm_exists},
@code{gdbm_fetch}, @code{gdbm_dump}, @code{gdbm_load},
@code{gdbm_export}, @code{gdbm_import}, @code{gdbm_reorganize},
@code{gdbm_firstkey}, @code{gdbm_nextkey}, @code{gdbm_store}.
@end defvr

@defvr {Error Code} GDBM_NO_DBNAME
Output database name is not specified.  This error code is set by
@code{gdbm_load} (@pxref{gdbm_load function,,gdbm_load}) if the first
argument points to @code{NULL} and the input file does not specify the
database name.
@end defvr

@defvr {Error Code} GDBM_ERR_FILE_OWNER
This error code is set by @code{gdbm_load} if it is unable to restore
database file owner.  It is a mild error condition, meaning that the
data have been restored successfully, only changing the target file
owner failed.  Inspect the system @code{errno} variable to get a more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_ERR_FILE_MODE
This error code is set by @code{gdbm_load} if it is unable to restore
database file mode.  It is a mild error condition, meaning that the data
have been restored successfully, only changing the target file owner
failed.  Inspect the system @code{errno} variable to get a more
detailed diagnostics.
@end defvr

@defvr {Error Code} GDBM_NEED_RECOVERY
Database is in inconsistent state and needs recovery.  Call
@code{gdbm_recover} if you get this error.  @xref{Recovery}, for a
detailed description of recovery functions.
@end defvr

@defvr {Error Code} GDBM_BACKUP_FAILED
The GDBM engine is unable to create backup copy of the file.
@end defvr

@defvr {Error Code} GDBM_DIR_OVERFLOW
Bucket directory would overflow the size limit during an attempt to split
hash bucket.  This error can occur while storing a new key.
@end defvr

@defvr {Error Code} GDBM_BAD_BUCKET
Invalid index bucket is encountered in the database.  Database
recovery is needed (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_BAD_HEADER
This error is set by @code{gdbm_open} and @code{gdbm_fd_open}, if the
first block read from the database file does not contain a valid
@command{GDBM} header.
@end defvr

@defvr {Error Code} GDBM_BAD_AVAIL
The available space stack is invalid.  This error can be set by
@code{gdbm_open} and @code{gdbm_fd_open}, if the extended database
verification was requested (@code{GDBM_XVERIFY}).  It is also set
by the @code{gdbm_avail_verify} function (@pxref{Additional
functions}).

Database recovery is needed (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_BAD_HASH_TABLE
Hash table in a bucket is invalid.  This error can be set by the
following functions: @code{gdbm_delete}, @code{gdbm_exists},
@code{gdbm_fetch}, @code{gdbm_firstkey}, @code{gdbm_nextkey}, and
@code{gdbm_store}.

Database recovery is needed (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_BAD_DIR_ENTRY
Bad directory entry found in the bucket.  The database recovery is
needed (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_FILE_CLOSE_ERROR
The @code{gdbm_close} function was unable to close the database file
descriptor.  The system @code{errno} variable contains the
corresponding error code.
@end defvr

@defvr {Error Code} GDBM_FILE_SYNC_ERROR
Cached content couldn't be synchronized to disk.  Examine the
@code{errno} variable to get more info,

Database recovery is needed (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_FILE_TRUNCATE_ERROR
File cannot be truncated.  Examine the @code{errno} variable to get
more info.

This error is set by @code{gdbm_open} and @code{gdbm_fd_open} when
called with the @code{GDBM_NEWDB} flag.
@end defvr

@defvr {Error Code} GDBM_BUCKET_CACHE_CORRUPTED
The bucket cache structure is corrupted.  Database recovery is needed
(@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_BAD_HASH_ENTRY
This error is set during sequential access (@pxref{Sequential}), if
the next hash table entry does not contain the expected key.  This
means that the bucket is malformed or corrupted and the database needs
recovery (@pxref{Recovery}).
@end defvr

@defvr {Error Code} GDBM_ERR_SNAPSHOT_CLONE
Set by the @code{gdbm_failure_atomic} function if it was unable to
clone the database file into a snapshot.  Inspect the system
@code{errno} variable for the underlying cause of the error.  If
@code{errno} is @code{EINVAL} or @code{ENOSYS}, crash tolerance
settings will be removed from the database.

@xref{Crash Tolerance API}.
@end defvr

@defvr {Error Code} GDBM_ERR_REALPATH
Set by the @code{gdbm_failure_atomic} function if the call to
@code{realpath} function failed.  @code{realpath} is used to
determine actual path names of the snapshot files.  Examine the system
@code{errno} variable for details.

@xref{Crash Tolerance API}.
@end defvr

@defvr {Error Code} GDBM_ERR_USAGE
Function usage error.  That includes invalid argument values, and the like.
@end defvr

@node Compatibility
@chapter Compatibility with standard @command{dbm} and @command{ndbm}

@cindex compatibility layer
@command{Gdbm} includes a compatibility layer, which provides traditional
@command{ndbm} and older @command{dbm} functions.  The layer is compiled and
installed if the @option{--enable-libgdbm-compat} option is used when
configuring the package.

@findex ndbm.h
@findex dbm.h
@findex libgdbm_compat
The compatibility layer consists of two header files: @file{ndbm.h}
and @file{dbm.h} and the @file{libgdbm_compat} library.

Older programs using @command{ndbm} or @command{dbm} interfaces can
use @file{libgdbm_compat} without any changes.  To link a program with
the compatibility library, add the following two options to the
@command{cc} invocation: @option{-lgdbm -lgdbm_compat}.  The @option{-L}
option may also be required, depending on where @command{GDBM} is
installed, e.g.:

@example
cc ... -lgdbm -lgdbm_compat
@end example

@cindex @samp{dir} file
@cindex @samp{pag} file
Databases created and manipulated by the compatibility interfaces
consist of two different files: @file{@var{file}.dir} and
@file{@var{file}.pag}.  This is required by the @acronym{POSIX}
specification and corresponds to the traditional usage.  Note,
however, that despite the similarity of the naming convention,
actual data stored in these files has not the same format as
in the databases created by other @command{dbm} or @command{ndbm}
libraries.  In other words, you cannot access a standard UNIX
@command{dbm} file with GNU @command{dbm}!

Compatibility interface includes only functions required by
@acronym{POSIX} (@pxref{ndbm}) or present in the traditional DBM
implementation (@pxref{dbm}).  Advanced @command{GDBM} features, such
as crash tolerance, cannot be used with such databases.

GNU @command{dbm} files are not @code{sparse}.  You can copy them with
the usual @code{cp} command and they will not expand in the copying
process.

@menu
* ndbm::  NDBM interface functions.
* dbm::   DBM interface functions.
@end menu

@node ndbm
@section NDBM interface functions
@cindex NDBM functions

The functions below implement the @acronym{POSIX} @command{ndbm} interface:

@deftypefn {ndbm} {DBM *} dbm_open (char *@var{file}, int @var{flags}, int @var{mode})
Opens a database.  The @var{file} argument is the full name of the
database file to be opened.  The function opens two files:
@file{@var{file}.pag} and @file{@var{file}.dir}.  The @var{flags} and
@var{mode} arguments have the same meaning as the second and third
arguments of @code{open} (@pxref{open,,,open(2), open(2) man page}),
except that a database opened for write-only access opens the files
for read and write access and the behavior of the @code{O_APPEND} flag is
unspecified.

The function returns a pointer to the @code{DBM} structure describing
the database.  This pointer is used to refer to this database in all
operations described below.

Any error detected will cause a return value of @code{NULL} and an
appropriate value will be stored in @code{gdbm_errno}
(@pxref{Variables}).
@end deftypefn

@deftypefn {ndbm} void dbm_close (DBM *@var{dbf})
Closes the database.  The @var{dbf} argument must be a pointer
returned by an earlier call to @code{dbm_open}.
@end deftypefn

@deftypefn {ndbm} datum dbm_fetch (DBM *@var{dbf}, datum @var{key})
Reads a record from the database with the matching key.  The @var{key}
argument supplies the key that is being looked for.

If no matching record is found, the @code{dptr} member of the returned
datum is @code{NULL}.  Otherwise, the @code{dptr} member of the
returned datum points to the memory managed by the compatibility
library.  The application should never free it.
@end deftypefn

@deftypefn {ndbm} int dbm_store (DBM *@var{dbf}, datum @var{key}, @
                                 datum @var{content}, int @var{mode})
Writes a key/value pair to the database.  The argument @var{dbf} is a
pointer to the @code{DBM} structure returned from a call to
@code{dbm_open}.  The @var{key} and @var{content} provide the values
for the record key and content.  The @var{mode} argument controls
the behavior of @code{dbm_store} in case a matching record already
exists in the database.  It can have one of the following two values:

@table @code
@kwindex DBM_REPLACE
@item DBM_REPLACE
Replace existing record with the new one.

@kwindex DBM_INSERT
@item DBM_INSERT
The existing record is left unchanged, and the function returns
@code{1}.
@end table

If no matching record exists in the database, new record will be
inserted no matter what the value of the @var{mode} is.
@end deftypefn

@deftypefn {ndbm} int dbm_delete (DBM *@var{dbf}, datum @var{key})
Deletes the record with the matching key from the database.  If the
function succeeds, @code{0} is returned.  Otherwise, if no matching
record is found or if an error occurs, @code{-1} is returned.
@end deftypefn

@deftypefn {ndbm} datum dbm_firstkey (DBM *@var{dbf})
Initializes iteration over the keys from the database and returns
the first key.  Note, that the word @samp{first} does not imply any
specific ordering of the keys.

If there are no records in the database, the @code{dptr} member of the
returned datum is @code{NULL}.  Otherwise, the @code{dptr} member of
the returned datum points to the memory managed by the compatibility
library.  The application should never free it.
@end deftypefn

@deftypefn {ndbm} datum dbm_nextkey (DBM *@var{dbf})
Continues the iteration started by @code{dbm_firstkey}.  Returns the
next key in the database.  If the iteration covered all keys in the
database, the @code{dptr} member of the returned datum is @code{NULL}.
Otherwise, the @code{dptr} member of the returned datum points to the
memory managed by the compatibility library.  The application should
never free it.

@cindex sequential access, using @samp{NDBM}
@cindex iteration loop, using @samp{NDBM}
The usual way of iterating over all the records in the database is:

@example
for (key = dbm_firstkey (dbf); key.ptr; key = dbm_nextkey (dbf))
  @{
    /* do something with the key */
  @}
@end example

The loop above should not try to delete any records from the database,
otherwise the iteration is not guaranteed to cover all the keys.
@xref{Sequential}, for a detailed discussion of this.
@end deftypefn

@deftypefn {ndbm} int dbm_error (DBM *@var{dbf})
Returns the error condition of the database: @code{0} if no errors
occurred so far while manipulating the database, and a non-zero value
otherwise.
@end deftypefn

@deftypefn {ndbm} void dbm_clearerr (DBM *@var{dbf})
Clears the error condition of the database.
@end deftypefn

@deftypefn {ndbm} int dbm_dirfno (DBM *@var{dbf})
Returns the file descriptor of the @samp{dir} file of the database.
It is guaranteed to be different from the descriptor returned by
the @code{dbm_pagfno} function (see below).

The application can lock this descriptor to serialize accesses to the
database.
@end deftypefn

@deftypefn {ndbm} int dbm_pagfno (DBM *@var{dbf})
Returns the file descriptor of the @samp{pag} file of the database.
See also @code{dbm_dirfno}.
@end deftypefn

@deftypefn {ndbm} int dbm_rdonly (DBM *@var{dbf})
Returns @code{1} if the database @var{dbf} is open in a read-only mode
and @code{0} otherwise.
@end deftypefn

@node dbm
@section DBM interface functions
@cindex DBM functions

The functions below are provided for compatibility with the old
UNIX @samp{DBM} interface.  Only one database at a time can be
manipulated using them.

@deftypefn {dbm} int dbminit (char *@var{file})
Opens a database.  The @var{file} argument is the full name of the
database file to be opened.  The function opens two files:
@file{@var{file}.pag} and @file{@var{file}.dir}.  If any of
them does not exist, the function fails.  It never attempts to create
the files.

The database is opened in the read-write mode, if its disk permissions
permit.

The application must ensure that the functions described below in
this section are called only after a successful call to @code{dbminit}.
@end deftypefn

@deftypefn {dbm} int dbmclose (void)
Closes the database opened by an earlier call to @code{dbminit}.
@end deftypefn

@deftypefn {dbm} datum fetch (datum @var{key})
Reads a record from the database with the matching key.  The @var{key}
argument supplies the key that is being looked for.

If no matching record is found, the @code{dptr} member of the returned
datum is @code{NULL}.  Otherwise, the @code{dptr} member of the
returned datum points to the memory managed by the compatibility
library.  The application should never free it.
@end deftypefn

@deftypefn {dbm} int store (datum @var{key}, datum @var{content})
Stores the key/value pair in the database.  If a record with the
matching key already exists, its content will be replaced with the new
one.

Returns @code{0} on success and @code{-1} on error.
@end deftypefn

@deftypefn {dbm} int delete (datum @var{key})
Deletes a record with the matching key.

If the function succeeds, @code{0} is returned.  Otherwise, if no
matching record is found or if an error occurs, @code{-1} is
returned.
@end deftypefn

@deftypefn {dbm} datum firstkey (void)
Initializes iteration over the keys from the database and returns
the first key.  Note, that the word @samp{first} does not imply any
specific ordering of the keys.

If there are no records in the database, the @code{dptr} member of the
returned datum is @code{NULL}.  Otherwise, the @code{dptr} member of
the returned datum points to the memory managed by the compatibility
library.  The application should never free it.
@end deftypefn

@deftypefn {dbm} datum nextkey (datum @var{key})
Continues the iteration started by a call to @code{firstkey}.  Returns
the next key in the database.  If the iteration covered all keys in the
database, the @code{dptr} member of the returned datum is @code{NULL}.
Otherwise, the @code{dptr} member of the returned datum points to the
memory managed by the compatibility library.  The application should
never free it.
@end deftypefn

@node gdbmtool
@chapter Examine and modify a GDBM database
@prindex gdbmtool

The @command{gdbmtool} utility allows you to view and modify an
existing @command{GDBM} database or to create a new one.

@cindex default database, @command{gdbmtool}
@flindex junk.gdbm
When invoked without arguments, it tries to open a database file called
@file{junk.gdbm}, located in the current working directory.  You can
change this default by supplying the name of the database as
argument to the program, e.g.:

@example
$ gdbmtool file.db
@end example

@cindex read-only mode, @command{gdbmtool}
@cindex @option{-r}, @command{gdbmtool} option
@cindex @option{--read-only}, @command{gdbmtool} option
The database will be opened in read-write mode, unless the
@option{-r} (@option{--read-only}) option is specified, in which case
it will be opened only for reading.

@cindex creating a database, @command{gdbmtool}
@cindex @option{-n}, @command{gdbmtool} option
@cindex @option{--newdb}, @command{gdbmtool} option
If the database does not exist, @command{gdbmtool} will create it.
There is a special option @option{-n} (@option{--newdb}), which
instructs the utility to create a new database.  If it is used and if
the database already exists, it will be deleted, so use it sparingly.

@menu
* invocation::
* shell::
@end menu

@node invocation
@section gdbmtool invocation
@cindex command line options, @command{gdbmtool}

When started without additional arguments, @command{gdbmtool} operates
on the default database @file{junk.gdbm}. Otherwise, the first
argument supplies the name of the database to operate upon. If neither
any additional arguments nor the @option{-f} (@option{--file}) option
are given, @command{gdbmtool} opens starts interactive shell and
receives commands directly from the human operator.

If more than one argument is given, all arguments past the database
name are parsed as @command{gdbmtool} commands (@pxref{shell}, for a
description of available commands) and executed in turn. All commands,
except the last one, should be terminated with semicolons. Semicolon
after the last command is optional. Note, that semicolons should be
escaped in order to prevent them from being interpreted by the shell.

Finally, if the @option{-f} (@option{--file}) option is supplied, its
argument specifies the name of the disk file with @command{gdbmtool}
script. The program will open that file and read commands from it.

The following table summarizes all @command{gdbmtool} command line
options:

@table @option
@item -b @var{size}
@itemx --block-size=@var{size}
Set block size.

@item -c @var{size}
@itemx --cache-size=@var{size}
Set cache size.

@item -d @var{fd}
@itemx --db-descriptor=@var{fd}
Use the database referred to by the file descriptor @var{fd}.  This
must be a valid open file descriptor, obtained by a call to
@code{open} (@pxref{open,,open a file,open(2), open(2) man page}),
@code{creat} or a similar function.  The database will be opened using
@code{gdbm_fd_open} (@pxref{gdbm_fd_open}).

This option is intended for use by automatic test suites.

@item -f @var{file}
@item --file @var{file}
Read commands from @var{file}, instead of the standard input.

@item -h
@itemx --help
Print a concise help summary.

@item -N
@itemx --norc
Don't read startup files (@pxref{startup files}).

@item -n
@itemx --newdb
Create the database.

@item -l
@itemx --no-lock
Disable file locking.

@item -m
@itemx --no-mmap
Disable memory mapping.

@item -T
@itemx --timing
Print time spent in each command.  This is equivalent to setting the
@code{timing} variable.  @xref{variables, timing}.

@item -t
@itemx --trace
Enable command tracing.  This is equivalent to setting the
@code{trace} variable.  @xref{variables, trace}.

@anchor{-q option}
@item -q
@itemx --quiet
Don't print the usual welcome banner at startup.  This is the same as
setting the variable @code{quiet} in the startup file.  @xref{quiet}.

@item -r
@itemx --read-only
Open the database in read-only mode.

@item -s
@itemx --synchronize
Synchronize to the disk after each write.

@item -V
@itemx --version
Print program version and licensing information and exit.

@item --usage
Print a terse invocation syntax summary along with a list of available
command line options.

@item -x
@itemx --extended
@itemx --numsync
Create new database in extended (numsync) format (@pxref{Numsync}).
This option sets the @code{format} variable to @samp{numsync}.
@xref{format variable}.
@end table

@node shell
@section gdbmtool interactive mode
@cindex interactive mode, @command{gdbmtool}

After successful startup, @command{gdbmtool} starts a loop, in which
it reads commands from the standard input, executes them and prints
results on the standard output.  If the standard input is attached
to a console, @command{gdbmtool} runs in interactive mode, which is
indicated by its @dfn{prompt}:

@example
gdbmtool> _
@end example

The utility finishes when it reads the @code{quit} command (see below) or
detects end-of-file on its standard input, whichever occurs first.

A @command{gdbmtool} command consists of a @dfn{command verb},
optionally followed by @dfn{arguments}, separated by any
amount of white space and terminated with a newline or semicolon.
A command verb can be entered either in full or in an abbreviated
form, as long as that abbreviation does not match any other verb.  For
example, @code{co} can be used instead of @code{count} and @code{ca}
instead of @code{cache}.

Any sequence of non-whitespace characters appearing after the command
verb forms an argument.  If the argument contains whitespace or
unprintable characters it must be enclosed in double quotes.  Within
double quotes the usual @dfn{escape sequences} are understood, as
shown in the table below:

@float Table, backslash-interpretation
@caption{Backslash escapes}
@multitable @columnfractions 0.30 .5
@item Sequence @tab Replaced with
@item \a @tab Audible bell character (@acronym{ASCII} 7)
@item \b @tab Backspace character (@acronym{ASCII} 8)
@item \f @tab Form-feed character (@acronym{ASCII} 12)
@item \n @tab Newline character (@acronym{ASCII} 10)
@item \r @tab Carriage return character (@acronym{ASCII} 13)
@item \t @tab Horizontal tabulation character (@acronym{ASCII} 9)
@item \v @tab Vertical tabulation character (@acronym{ASCII} 11)
@item \\ @tab Single slash
@item \" @tab Double quote
@end multitable
@end float

In addition, a backslash immediately followed by the end-of-line
character effectively removes that character, allowing to split long
arguments over several input lines.

Command parameters may be optional or mandatory.  If the number of
actual arguments is less than the number of mandatory parameters,
@command{gdbmtool} will prompt you to supply missing arguments.  For
example, the @code{store} command takes two mandatory parameters, so
if you invoked it with no arguments, you would be prompted twice to
supply the necessary data, as shown in  example below:

@example
gdbmtool> @kbd{store}
key? @kbd{three}
data? @kbd{3}
@end example

However, such prompting is possible only in interactive mode.  In
non-interactive mode (e.g.@: when running a script), all arguments must
be supplied with each command, otherwise @command{gdbmtool} will report an
error and exit immediately.

@cindex readline
@cindex GNU Readline
If the package is compiled with GNU Readline, the input line can be
edited (@pxref{Command Line Editing, ,
Command Line Editing, readline, GNU Readline Library}).

@menu
* variables::      shell variables.
* commands::       shell commands.
* definitions::    how to define structured data.
* startup files::
@end menu

@node variables
@subsection Shell Variables
@cindex variables, gdbmtool
A number of @command{gdbmtool} parameters is kept in its internal
variables.  To examine or modify variables, use the @code{set} command
(@pxref{set}).

@deftypevr {gdbmtool variable} bool confirm
Whether to ask for confirmation before certain destructive operations,
such as truncating the existing database.

Default is @code{true}.
@end deftypevr

@deftypevr {gdbmtool variable} string delim1
A string used to delimit fields of a structured datum on output
(@pxref{definitions}).

Default is @samp{,} (a comma).  This variable cannot be unset.
@end deftypevr

@deftypevr {gdbmtool variable} string delim2
A string used to delimit array items when printing a structured datum
(@pxref{definitions}).

Default is @samp{,} (a comma).  This variable cannot be unset.
@end deftypevr

@deftypevr {gdbmtool variable} string errorexit
@deftypevrx {gdbmtool variable} bool errorexit
Comma-delimited list of @command{GDBM} error codes which cause program
termination.  Error codes are specified via their canonical names
(@pxref{Error codes}).  The @code{GDBM_} prefix can be omitted.  Code
name comparison is case-insensitive.  Each error code can optionally
be prefixed with minus sign, to indicate that it should be removed
from the resulting list, or with plus sign (which is allowed for
symmetry).  A special code @samp{all} stands for all available error codes.

In boolean context, the @code{true} value is equivalent to @samp{all},
and @code{false} (i.e. variable unset) is equivalent to @samp{-all}.

This variable cannot be set from interactive sessions.
@end deftypevr

@deftypevr {gdbmtool variable} string errormask
@deftypevrx {gdbmtool variable} bool errormask
Comma-delimited list of @command{GDBM} error codes which are masked, i.e.
which won't trigger a diagnostic message if they occur.  The syntax is
the same as described for @code{errorexit}.
@end deftypevr

@deftypevr {gdbmtool variable} string pager
The name and command line of the pager program to pipe output to.
This program is used in interactive mode when the estimated number of
output lines is greater then the number of lines on your screen.

The default value is inherited from the environment variable
@env{PAGER}.  Unsetting this variable disables paging.
@end deftypevr

@deftypevr {gdbmtool variable} string ps1
Primary prompt string.  Its value can contain @dfn{conversion
specifiers}, consisting of the @samp{%} character followed by another
character.  These specifiers are expanded in the resulting prompt as
follows:

@multitable @columnfractions 0.4 0.5
@headitem Sequence @tab Expansion
@item %f @tab name of the current database file
@item %p @tab program invocation name
@item %P @tab package name (@samp{GDBM})
@item %v @tab program version
@item %_ @tab single space character
@item %% @tab %
@end multitable

The default value is @samp{%p>%_}, i.e. the program name, followed by
a ``greater than'' sign, followed by a single space.
@end deftypevr

@deftypevr {gdbmtool variable} string ps2
Secondary prompt.  See @code{ps1} for a description of its value.
This prompt is displayed before reading the second and subsequent
lines of a multi-line command.

The default value is @samp{%_>%_}.
@end deftypevr

@deftypevr {gdbmtool variable} bool timing
When each command terminates, print an additional line listing times
spent in that command.  The line is formatted as follows:

@example
[reorganize r=0.070481 u=0.000200 s=0.000033]
@end example

@noindent
Here, @samp{reorganize} is the name of the command that finished, the
number after @samp{r=} is real time spent executing the command, the
number after @samp{u=} is the user CPU time used and the number after
@samp{s=} is the system CPU time used.
@end deftypevr

@deftypevr {gdbmtool variable} bool trace
Enable command tracing.  This is similar to the shell @option{-t}
option: before executing each command, @command{gdbmtool} will print
on standard error a line starting with a plus sign and followed by the
command name and its arguments.
@end deftypevr

@anchor{quiet}
@deftypevr {gdbmtool variable} bool quiet
Whether to display a welcome banner at startup.  To affect
@command{gdbmtool}, this variable should be set in a startup script
file (@pxref{startup files}).  @xref{-q option}.
@end deftypevr

@anchor{open parameters}
The following variables control how the database is opened:

@deftypevr {gdbmtool variable} numeric blocksize
Sets the block size.  @xref{Open, block_size}.  Unset by default.
@end deftypevr

@deftypevr {gdbmtool variable} numeric cachesize
Sets the cache size.  @xref{Options, GDBM_SETCACHESIZE}.

This variable affects the currently opened database immediately.  It
is also used by @command{open} command.

To enable automatic cache size selection, unset this variable.  This
is the default.
@end deftypevr

@deftypevr {gdbmtool variable} string filename
Name of the database file.  If the @code{open} command is called
without argument (e.g. called implicitly), this variable names the
database file to open.  If @code{open} is called with file name
argument, upon successful opening of the database the @code{filename}
variable is initialized with its file name.

This variable cannot be unset.
@end deftypevr

@deftypevr {gdbmtool variable} number fd
File descriptor of the database file to open.  If this variable is
set, its value must be an open file descriptor referring to a
@command{GDBM} database file.  The @code{open} command will use
@code{gdbm_fd_open} function to use this file (@pxref{gdbm_fd_open}).
When this database is closed, the descriptor will be closed as well
and the @code{fd} variable will be unset.

See also the @option{-d} (@option{--db-descriptor}) command line
option in @ref{invocation}.
@end deftypevr

@anchor{format variable}
@deftypevr {gdbmtool variable} string format
Defines the format in which new databases will be created.  Allowed
values are:

@table @samp
@item standard
Databases will be created in standard format.  This is the format used
by all @command{GDBM} versions prior to 1.21.  This value is the
default.

@item numsync
Extended format, best for crash-tolerant applications.
@xref{Numsync}, for a discussion of this format.
@end table

@end deftypevr

@anchor{openvar}
@deftypevr {gdbmtool variable} string open
Open mode.  The following values are allowed:

@table @asis
@item newdb
Truncate the database if it exists or create a new one.  Open it in
read-write mode.

Technically, this sets the @code{GDBM_NEWDB} flag in call to @code{gdbm_open}.
@xref{Open, GDBM_NEWDB}.
@item wrcreat
@itemx rw
Open the database in read-write mode.  Create it if it does not
exist.  This is the default.

Technically speaking, it sets the @code{GDBM_WRCREAT} flag in call to
@code{gdbm_open}.  @xref{Open, GDBM_WRCREAT}.
@item reader
@itemx readonly
Open the database in read-only mode.  Signal an error if it does not
exist.

This sets the @code{GDBM_READER} flag (@pxref{Open, GDBM_READER}).
@end table

Attempting to set any other value or to unset this variable results
in error.
@end deftypevr

@anchor{filemode}
@deftypevr {gdbmtool variable} number filemode
File mode (in octal) for creating new database files and database
dumps.
@end deftypevr

@deftypevr {gdbmtool variable} bool lock
Lock the database.  This is the default.

Setting this variable to false or unsetting it results in passing
@code{GDBM_NOLOCK} flag to @code{gdbm_open} (@pxref{Open, GDBM_NOLOCK}).
@end deftypevr

@deftypevr {gdbmtool variable} bool mmap
Use memory mapping.  This is the default.

Setting this variable to false or unsetting it results in passing
@code{GDBM_NOMMAP} flag to @code{gdbm_open} (@pxref{Open, GDBM_NOMMAP}).
@end deftypevr

@deftypevr {gdbmtool variable} bool sync
Flush all database writes on disk immediately.  Default is false.
@xref{Open, GDBM_SYNC}.
@end deftypevr

@deftypevr {gdbmtool variable} bool coalesce
Enables the @emph{coalesce} mode, i.e. merging of the freed blocks of
@command{GDBM} files with entries in available block lists. This
provides for effective memory management at the cost of slight
increase in execution time when calling
@code{gdbm_delete}. @xref{Options, GDBM_SETCOALESCEBLKS}.

This variable affects the currently opened database immediately and
will be used by @command{open} command, when it is invoked.
@end deftypevr

@deftypevr {gdbmtool variable} bool centfree
Set to @code{true}, enables the use of central free block pool in
newly opened databases. @xref{Options, GDBM_SETCENTFREE}.

This variable affects the currently opened database immediately and
will be used by @command{open} command, when it is invoked.
@end deftypevr

The following commands are used to list or modify the variables:

@anchor{set}
@deffn {command verb} set [@var{assignments}]
When used without arguments, lists all variables and their values.
Unset variables are shown after a comment sign (@samp{#}).  For string
and numeric variables, values are shown after an equals sign.  For
boolean variables, only the variable name is displayed if the variable
is @code{true}.  If it is @code{false}, its name is prefixed with
@samp{no}.

For example:

@example
@group
# blocksize is unset
# cachesize is unset
nocentfree
nocoalesce
confirm
delim1=","
delim2=","
# fd is unset
filemode=644
filename="junk.gdbm"
format="standard"
lock
mmap
open="wrcreat"
pager="less"
ps1="%p>%_"
ps2="%_>%_"
# quiet is unset
nosync
@end group
@end example

If used with arguments, the @code{set} command alters the specified
variables.  In this case, arguments are variable assignments in the
form @samp{@var{name}=@var{value}}.  For boolean variables, the
@var{value} is interpreted as follows: if it is numeric, @code{0}
stands for @code{false}, any non-zero value stands for @code{true}.
Otherwise, the values @code{on}, @code{true}, and @code{yes} denote
@code{true}, and @code{off}, @code{false}, @code{no} stand for
@code{false}.  Alternatively, only the name of a boolean variable can be
supplied to set it to @code{true}, and its name prefixed with
@code{no} can be used to set it to false.  For example, the following
command sets the @code{delim2} variable to @samp{;} and the
@code{confirm} variable to @code{false}:

@example
set delim2=";" noconfirm
@end example
@end deffn

@deffn {command verb} unset @var{variables}
Unsets the listed variables.  The effect of unsetting depends on the
variable.  Unless explicitly described in the discussion of the
variables above, unsetting a boolean variable is equivalent to setting it to
@code{false}.  Unsetting a string variable is equivalent to assigning it
an empty string.
@end deffn

@node commands
@subsection Gdbmtool Commands

@deffn {command verb} avail
Print the @dfn{avail list}.
@end deffn

@deffn {command verb} bucket @var{num}
Print the bucket number @var{num} and set it as the current one.
@end deffn

@deffn {command verb} cache
Print the bucket cache.
@end deffn

@deffn {command verb} close
Close the currently open database.
@end deffn

@deffn {command verb} count
Print the number of entries in the database.
@end deffn

@deffn {command verb} current
Print the current bucket.
@end deffn

@deffn {command verb} debug [[+-]@var{token}...]
If @command{GDBM} is configured with additional debugging, this
statement queries or sets @command{GDBM} internal debugging level.
This is intended for debugging and testing purposes and requires
good knowledge of @command{GDBM} internals. The use of this command is
not recommended.
@end deffn

@deffn {command verb} delete @var{key}
Delete record with the given @var{key}
@end deffn

@deffn {command verb} dir
Print hash directory.
@end deffn

@deffn {command verb} downgrade
Downgrade the database from extended to the standard database format.
@xref{Numsync}.
@end deffn

@anchor{gdbmtool export}
@deffn {command verb} export @var{file-name} [truncate] [binary|ascii]
Export the database to the flat file @var{file-name}.  @xref{Flat files},
for a description of the flat file format and its purposes.  This
command will not overwrite an existing file, unless the
@samp{truncate} parameter is also given.  Another optional argument
determines the type of the dump (@pxref{Flat files}).  By default, ASCII
dump is created.

The global variable @code{filemode} specifies the permissions to use
for the created output file.
@end deffn

@deffn {command verb} fetch @var{key}
Fetch and display the record with the given @var{key}.
@end deffn

@deffn {command verb} first
Fetch and display the first record in the database.  Subsequent
records can be fetched using the @code{next} command (see below).
@xref{Sequential}, for more information on sequential access.
@end deffn

@deffn {command verb} hash @var{key}
Compute and display the hash value for the given @var{key}.
@end deffn

@deffn {command verb} header
Print file header.
@end deffn

@deffn {command verb} help
@deffnx {command verb} ?
Print a concise command summary, showing each command verb
with its parameters and a short description of what it does.  Optional
arguments are enclosed in square brackets.
@end deffn

@anchor{gdbmtool import}
@deffn {command verb} import @var{file-name} [replace] [nometa]
Import data from a flat dump file @var{file-name}
(@pxref{Flat files}).  If the word @samp{replace} is given
as an argument, any records with the same keys as the already
existing ones will replace them.  The word @samp{nometa} turns off
restoring meta-information from the dump file.
@end deffn

@deffn {command verb} history
@deffnx {command verb} history @var{count}
@deffnx {command verb} history @var{n} @var{count}
Shows the command history list with line numbers.  When used without
arguments, shows entire history.  When used with one argument,
displays @var{count} last commands from the history.  With two
arguments, displays @var{count} commands starting from @var{n}th
command.  Command numbering starts with 1.

This command is available only if @command{GDBM} was compiled with GNU
Readline.  The history is saved in file @file{.gdbmtool_history} in
the user's home directory.  If this file exists upon startup, it is read to
populate the history.  Thus, command history is preserved between
@command{gdbmtool} invocations.
@end deffn

@deffn {command verb} list
List the contents of the database.
@end deffn

@deffn {command verb} next [@var{key}]
Sequential access: fetch and display the next record.  If the @var{key} is
given, the record following the one with this key will be fetched.

Issuing several @code{next} commands in row is rather common.  A
shortcut is provided to facilitate such use: if the last entered
command was @command{next}, hitting the @kbd{Enter} key repeats it
without arguments.

See also @code{first}, above.

@xref{Sequential}, for more information on sequential access.
@end deffn

@deffn {command verb} open @var{filename}
@deffnx {command verb} open
Open the database file @var{filename}.  If used without arguments, the
database name is taken from the variable @code{filename}.

If successful, any previously open database is closed and the
@code{filename} variable is updated.  Otherwise, if the operation
fails, the currently opened database remains unchanged.

This command takes additional information from the following
variables:

@table @code
@item filename
Name of the database to open, if no argument is given.

@item fd
File descriptor to use.  If set, this must be an open file descriptor
referring to a valid database file.  The database will be opened using
@code{gdbm_fd_open} (@pxref{gdbm_fd_open}).  The file descriptor will
be closed and the variable unset upon closing the database.

@item filemode
Specifies the permissions to use in case a new file is created.

@item open
The database access mode.  @xref{openvar,, The @var{open} variable},
for a list of its values.

@item lock
Whether or not to lock the database.  Default is @code{on}.

@item mmap
Use the memory mapping.  Default is @code{on}.

@item sync
Synchronize after each write.  Default is @code{off}.

@end table

@xref{open parameters}, for a detailed description of these variables.
@end deffn

@deffn {command verb} perror [@var{code}]
Describe the given @command{GDBM} error code.

The description occupies one or two lines.  The second line is present
if the system error number should be checked when handling this code.
In this case, the second line states @samp{Examine errno}.

If @var{code} is omitted, the latest error that occurred in the
current database is described.  Second line of the output (if
present), contains description of the latest system error.

Example:

@example
gdbmtool> perror 3
GDBM error code 3: "File open error"
Examine errno.
@end example
@end deffn

@deffn {command verb} quit
Close the database and quit the utility.
@end deffn

@deffn {command verb} recover [@var{options}]
Recover the database from structural inconsistencies.  @xref{Database
consistency}.

The following @var{options} are understood:

@table @option
@item backup
Create a backup copy of the original database.

@item max-failed-buckets=@var{n}
Abort recovery process if @var{n} buckets could not be recovered.

@item max-failed-keys=@var{n}
Abort recovery process if @var{n} keys could not be recovered.

@item max-failures=@var{n}
Abort recovery process after @var{n} failures. A @dfn{failure} in this
context is either a key or a bucket that failed to be recovered.

@item summary
Print the recovery statistics at the end of the run. The statistics
includes number of successfully recovered, failed and duplicate keys
and the number of recovered and failed buckets.

@item verbose
Verbosely list each error encountered.
@end table
@end deffn

@deffn {command verb} reorganize
Reorganize the database (@pxref{Reorganization}).
@end deffn

@deffn {command verb} shell @var{command}
@deffnx {command verb} ! @var{command}
Execute @var{command} via current shell.  If @var{command} is empty,
shell is started without additional arguments.  Otherwise, it is run
as @samp{$SHELL -c @var{command}}.

For convenience, @var{command} is not parsed as @command{gdbmtool}
command line.  It is passed to the shell verbatim.  It can include
newline characters if these are preceded by a backslash or appear
within singly or doubly quoted strings.

When using @code{!} form, be sure to separate it from @var{command} by
whitespace, otherwise it will be treated as readline @dfn{event specifier}.
@end deffn

@deffn {command verb} snapshot @var{filename} @var{filename}
Analyze two snapshot files and select the most recent of them.  In
case of error, display a detailed diagnostics and meta-information of
both snapshots.

@xref{Manual crash recovery}, for a detailed discussion.
@end deffn

@deffn {command verb} source @var{filename}
Read @command{gdbmtool} commands from the file @var{filename}.
@end deffn

@deffn {command verb} status
Print current program status.  The following example shows the
information displayed:

@example
Database file: junk.gdbm
Database is open
define key string
define content string
@end example

The two @code{define} strings show the defined formats for key and
content data.  @xref{definitions}, for a detailed discussion of their
meaning.
@end deffn

@deffn {command verb} store @var{key} @var{data}
Store the @var{data} with @var{key} in the database.  If @var{key}
already exists, its data will be replaced.
@end deffn

@deffn {command verb} sync
Synchronize the database with the disk storage (@pxref{Sync}).
@end deffn

@deffn {command verb} upgrade
Upgrade the database from standard to extended database format.
@xref{Numsync}.
@end deffn

@deffn {command verb} version
Print the version of @command{gdbm}.
@end deffn

@node definitions
@subsection Data Definitions
@command{GDBM} databases are able to keep data of any type, both in the key and
in the content part of a record.  Quite often these data are
structured, i.e. they consist of several fields of various types.
@command{Gdbmtool} provides a mechanism for handling such kind of
records.

The @code{define} command defines a record structure.  The general
syntax is:

@example
define @var{what} @var{definition}
@end example

@noindent
where @var{what} is @code{key} to defining the structure of key data and
@code{content} to define the structure of the content records.

The @var{definition} can be of two distinct formats.  In the simplest
case it is a single data type.  For example,

@example
define content int
@end example

@noindent
defines content records consisting of a single integer field.
Supported data types are:

@table @asis
@item char
Single byte (signed).
@item short
Signed short integer.
@item ushort
Unsigned short integer.
@item int
Signed integer.
@item unsigned
@itemx uint
Unsigned integer.
@item long
Signed long integer.
@item ulong
Unsigned long integer.
@item llong
Signed long long integer.
@item ullong
Unsigned long long integer.
@item float
A floating point number.
@item double
Double-precision floating point number.
@item string
Array of bytes.
@item stringz
Null-terminated string, trailing null being part of the string.
@end table

All numeric data types (integer as well as floating point) have the
same respective widths as in C language on the host where the database
file resides.

The @code{string} and @code{stringz} are special.  Both define a
string of bytes, similar to @samp{char x[]} in C.  The former
defines an array of bytes, the latter - a null-terminated string.
This makes a difference, in particular, when the string is the only
part of datum.  Consider the following two definitions:

@enumerate 1
@item @code{define key string}
@item @code{define key stringz}
@end enumerate

@noindent
Now, suppose we want to store the string "ab" in the key.  Using the
definition (1), the @code{dptr} member of @command{GDBM} @code{datum} will
contain two bytes: @samp{a}, and @samp{b}.  Consequently, the
@code{dsize} member will have the value 2.  Using the definition (2),
the @code{dptr} member will contain three bytes: @samp{a}, @samp{b},
and ASCII 0.  The @code{dsize} member will have the value 3.

The definition (1) is the default for both key and content.

The second form of the @code{define} statement is similar to the C
@code{struct} statement and allows for defining structural data.   In
this form, the @var{definition} part is a comma-separated list of data
types and variables enclosed in curly braces.  In contrast to the
rest of @command{gdbm} commands, this command is inherently
multiline and is terminated with the closing curly brace.  For
example:

@example
define content @{
        int status,
        pad 8,
        char id[3],
        string name
@}
@end example

@noindent
This defines a structure consisting of three members: an integer
@code{status}, an array of 3 bytes @code{id}, and an array of
bytes @code{name}.  Notice the @code{pad} statement: it allows to
introduce padding between structure members.  Another useful statement
is @code{offset}: it specifies that the member following it begins at
the given offset in the structure.  Assuming the size of @code{int} is
8 bytes, the above definition can also be written as

@example
define content @{
        int status,
        offset 16,
        char id[3],
        string name
@}
@end example

@emph{NOTE}: The @code{string} type can reasonably be used only if it
is the last or the only member of the data structure.  That's because it
provides no information about the number of elements in the array, so
it is interpreted to contain all bytes up to the end of the datum.

When displaying the structured data, @command{gdbmtool} precedes each
value with the corresponding field name and delimits parts of the
structure with the string defined in the @code{delim1} variable
(@pxref{variables}).  Array elements are delimited using the string from
@code{delim2}.  For example:

@example
gdbmtool> fetch foo
status=2,id=@{ a, u, x @},name="quux"
@end example

To supply a structured datum as an argument to a @command{gdbmtool}
command, use the same notation, e.g.:

@example
gdbmtool> store newkey @{ status=2, id=@{a,u,x@}, name="quux" @}
@end example

The order in which the fields are listed is not significant.  The
above command can as well be written as:

@example
gdbmtool> store newkey @{ id=@{a,u,x@}, status=2, name="quux" @}
@end example

You are not required to supply all defined fields.  Any number of them
can be omitted, provided that at least one remains.  The omitted
fields are filled with 0:

@example
gdbmtool> store newkey @{ name="bar" @}
gdbmtool> fetch newkey
status=0,id=@{ ,, @},name=bar
@end example

Yet another way to supply structured data to a command is by listing
the value for each field in the order they are defined, without field
names:

@example
gdbmtool> store newkey @{ 2, @{a,u,x@}, "quux" @}
@end example

@node startup files
@subsection Startup Files
@cindex startup file, gdbmtool
@cindex init file, gdbmtool
@flindex .gdbmtoolrc
Upon startup @command{gdbmtool} looks for a file named
@file{.gdbmtoolrc} first in the current working directory and, if not
found, in the home directory of the user who started the command.

If found, this file is read and interpreted as a list of
@command{gdbmtool} commands.  This allows you to customize the
program behavior.

Following is an example startup file which disables the welcome
banner, sets command line prompt to contain the name of the database
file in parentheses and defines the structure of the database content
records:

@example
@group
set quiet
set ps1="(%f) "
define key stringz
define content @{
        int time,
        pad 4,
        int status
@}
@end group
@end example

@node gdbm_dump
@chapter The @command{gdbm_dump} utility
@prindex gdbm_dump

The @command{gdbm_dump} utility creates a flat file dump of a @command{GDBM}
database (@pxref{Flat files}).  It takes one mandatory argument: the
name of the source database file.  The second argument, if given,
specifies the name of the output file.  If not given,
@command{gdbm_dump} will produce the dump on the standard output.

For example, the following invocation creates a dump of the database
@file{file.db} in the file @file{file.dump}:

@example
$ gdbm_dump file.db file.dump
@end example

By default the utility creates dumps in ASCII format (@pxref{Flat
files,ASCII}).  Another format can be requested using the
@option{--format} (@option{-H}) option.

The @command{gdbm_dump} utility understands the following command line
options:

@table @option
@item -H @var{fmt}
@itemx --format=@var{fmt}
Select output format.  Valid values for @var{fmt} are: @code{binary}
or @code{0} to select binary dump format, and @code{ascii} or @code{1}
to select ASCII format.

@item -h
@itemx --help
Print a concise help summary.

@item -V
@itemx --version
Print program version and licensing information and exit.

@item --usage
Print a terse invocation syntax summary along with a list of available
command line options.
@end table

@node gdbm_load
@chapter The @command{gdbm_load} utility
@prindex gdbm_load

The @command{gdbm_load} utility restores a @command{GDBM} database from a flat
file.  The utility requires at least one argument: the name of the
input flat file.  If it is @samp{-}, the standard input will be read.
The format of the input file is detected automatically.

By default the utility attempts to restore the database under its
original name, as stored in the input file.  It will fail to do so if
the input is in binary format.  In that case, the name of the database
must be given as the second argument.

In general, if two arguments are given, the second one is treated as
the name of the database to create, overriding the file name specified
in the flat file.  All existing keys will be removed from this
database prior to loading from the dump.  Use the @option{--update}
(@option{-U}) option if it is not what you wish.

When given the @option{--update} (@option{-U}) option,
@command{gdbm_load} will update the existing database with the data
from the dump.  It will bail out if the dump contains a key that is
already present in the database.  To silently overwrite existing keys,
use the @option{--replace} (@option{-r}) option.

The utility understands the following command line options:

@table @option

@item -b @var{num}
@itemx --block-size=@var{num}
Sets block size.  @xref{Open, block_size}.

@item -c @var{num}
@itemx --cache-size=@var{num}
Sets cache size.  @xref{Options, GDBM_SETCACHESIZE}.

@item -M
@itemx --mmap
Use memory mapping.

@item -m @var{mode}
@item --mode=@var{mode}
Sets the file mode.  The argument is the desired file mode in octal.

@item -n
@itemx --no-meta
Do not restore file meta-data (ownership and mode) from the flat file.

@item -r
@itemx --replace
Replace existing keys.  This option can be used only together with
@option{--update} (@option{-U}).

@item -U
@itemx --update
Update an existing database.  The database name must be given in the
second argument to @command{gdbm_load}.  The key/value pairs from the
dump file will be added to that database, without removing the
existing keys.  To overwrite the existing keys from the dump file, use
@option{--update --replace}.

@item -u @var{owner}[:[@var{group}]]
@itemx --user=@var{owner}[:[@var{group}]]
Set file owner.  The @var{owner} can be either a valid user name or
UID.  Similarly, the @var{group} is either a valid group name or GID.
If @var{group} is not given, the main group of @var{owner} is implied, if
@var{owner} is followed by a @samp{:}, otherwise the login group of the
current user is implied.

User and group parts can be separated by a dot, instead of the colon,
but such usage is discouraged.

@item -h
@itemx --help
Print a concise help summary.

@item -V
@itemx --version
Print program version and licensing information and exit.

@item --usage
Print a terse invocation syntax summary along with a list of available
command line options.
@end table

@node Exit codes
@chapter Exit codes
@cindex exit code

All @command{GDBM} utilities return uniform exit codes.  These are
summarized in the table below:

@multitable @columnfractions 0.3 0.7
@headitem Code @tab Meaning
@item 0 @tab Successful termination.
@item 1 @tab A fatal error occurred.
@item 2 @tab Program was unable to restore file ownership or mode.
@item 3 @tab Command line usage error.
@end multitable

@node Bugs
@chapter Problems and bugs

If you have problems with GNU @code{dbm} or think you've found a bug,
please report it.  Before reporting a bug, make sure you've actually
found a real bug.  Carefully reread the documentation and see if it
really says you can do what you're trying to do.  If it's not clear
whether you should be able to do something or not, report that too; it's
a bug in the documentation!

Before reporting a bug or trying to fix it yourself, try to isolate it
to the smallest possible input file that reproduces the problem.  Then
send us the input file and the exact results @command{GDBM} gave you.  Also
say what you expected to occur; this will help us decide whether the
problem was really in the documentation.

Once you've got a precise problem, send e-mail to
@email{bug-gdbm@@gnu.org}.

Please include the version number of GNU @code{dbm} you are using.  You can get
this information by printing the variable @code{gdbm_version}
(@pxref{Variables}).

Non-bug suggestions are always welcome as well.  If you have questions
about things that are unclear in the documentation or are just obscure
features, please report them too.

You may contact the authors and maintainers by e-mail:
Philip Nelson @email{phil@@cs.wwu.edu},
Jason Downs @email{downsj@@downsj.com},
Sergey Poznyakoff @email{gray@@gnu.org} or @email{gray@@gnu.org.ua}.

Crash tolerance support written by Terence Kelly
@email{tpkelly@@acm.org}, @email{tpkelly@@cs.princeton.edu}, or
@email{tpkelly@@eecs.umich.edu}.

@node Resources
@chapter Additional resources

For the latest updates and pointers to additional resources, visit
@uref{http://www.gnu.org/@/software/@/gdbm}.

In particular, a copy of @command{GDBM} documentation in various formats
is available online at @uref{http://www.gnu.org/@/software/@/gdbm/@/manual.html}.

Latest versions of @command{GDBM} can be downloaded from anonymous FTP:
@uref{ftp://ftp.gnu.org/@/gnu/@/gdbm}, or via HTTP from
@uref{http://ftp.gnu.org/@/gnu/@/gdbm}, or via HTTPS from
@uref{https://ftp.gnu.org/@/gnu/@/gdbm},
or from any
@ifhtml
@uref{http://www.gnu.org/order/ftp.html,,GNU mirror} worldwide.
@end ifhtml
@ifnothtml
GNU mirror worldwide.  See @uref{http://www.gnu.org/@/order/@/ftp.html},
for a list of mirrors.
@end ifnothtml

To track @command{GDBM} development, visit
@uref{http://puszcza.gnu.org.ua/@/projects/@/gdbm}.

@node GNU Free Documentation License
@appendix GNU Free Documentation License

@include fdl.texi

@node Index
@unnumbered Index

@printindex cp

@ifset WEBDOC
@ifhtml
@node This Manual in Other Formats
@unnumbered This Manual in Other Formats
@include otherdoc.texi
@end ifhtml
@end ifset

@bye