summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* added a coding conventions documentAdrian Thurston2017-01-021-0/+66
|
* minor fixes: a missing cast, no comma at enum end, no redefined typedefsPeter Reijnders2016-12-197-21/+10
|
* correct name of realm names var, use var instead of stringsPeter Reijnders2016-12-193-11/+11
|
* added autoconf and automake code for building asciidoc manualAdrian Thurston2016-12-194-1/+39
|
* docs: split code from prosePeter Reijnders2016-12-1919-184/+212
|
* rough documentation outlinePeter Reijnders2016-12-1931-0/+714
|
* fixed type of strtol return val in lexFactorRepNumAdrian Thurston2016-12-191-1/+1
| | | | Patch from Peter Reijnders
* fixed class vs struct inconsistency for FsmCodeGenAdrian Thurston2016-12-191-1/+1
|
* usage improvements from Peter ReijndersAdrian Thurston2016-12-191-2/+12
|
* migrated from GPLv2 to MIT licenseAdrian Thurston2016-12-13122-2719/+2178
|
* an automated include convention improvement from Peter ReijndersAdrian Thurston2016-12-1377-434/+392
|
* call ldconfig in a postinst, since we now export a shared libAdrian Thurston2016-12-051-0/+6
|
* if an or block returns null then return lambda fsmAdrian Thurston2016-10-291-0/+4
|
* removed clones of cons-item-list resolveAdrian Thurston2016-10-292-37/+23
|
* some name improvements in colm grammarAdrian Thurston2016-10-012-34/+34
|
* auto-trim in print, send and construct, disable with @Adrian Thurston2016-10-018-11/+70
| | | | | | | | | | | | | | Now automatically trimming when building trees and printing. This is consistent with the $ operator. Rationale is to standardize on the most commonly desired behaviour and provide options for the less-needed. Also much less likely to result in surprising output. There are many cases where test inputs won't have any surrounding ignores, making it easy to forget a trim is required. Then the occasional tree with surrounding ignores will give an undesired result. Better to standardize on trim, making the default behaviour more consistent. The trim can be turned off with the @ operator in front of an expression.
* if matching a ref, return a tree typeAdrian Thurston2016-09-231-0/+5
| | | | A match will turn a ref into a tree. Need to return a tree type.
* added colm_error for extracting the error stringAdrian Thurston2016-08-232-0/+13
|
* added a function for retrieving the reducer contextAdrian Thurston2016-08-012-0/+6
|
* test suite removed from colm.git, now maintained in test-colmAdrian Thurston2016-07-31168-27949/+0
|
* implemented a const-string define option -Ename=valAdrian Thurston2016-07-288-53/+180
|
* check include paths when opening includesAdrian Thurston2016-07-281-2/+15
|
* test suite: link to proper libcolm and set LD_LIBRARY_PATHAdrian Thurston2016-07-243-4/+6
|
* version bump to 0.13.0.4colm-0.13.0.4Adrian Thurston2016-07-111-2/+2
|
* main needs to use -lcolm nowAdrian Thurston2016-07-061-2/+2
|
* don't build 'p' and 'd' versions of the lib (production & debug), just libcolmAdrian Thurston2016-06-292-13/+8
|
* fix a spello which caused a compilation failureJan Engelhardt2016-06-291-1/+1
|
* build libcolm shared and use -release ld optionJan Engelhardt2016-06-292-9/+12
|
* commit shift count needs to be initialized to -1Adrian Thurston2016-03-232-0/+12
| | | | | An initial value of zero for commit shift count means we can erroneously fail a parse when we back up to zero shifts.
* use tree_child in the reduction commitAdrian Thurston2016-03-232-2/+2
| | | | Otherwise ignores and attributes will put the walks out of sync.
* removed stream name transfer hack that was necessary before managed namesAdrian Thurston2016-03-171-7/+3
| | | | | Removed name transfer hack that is no longer necessary and now is a bug. Neglected to remove this when the name management was added.
* extract stream file names so we can use them after the colm program is deletedAdrian Thurston2016-03-164-0/+16
|
* added missing free in system functionAdrian Thurston2016-03-101-1/+3
|
* manage file name allocations and allow them to be exportedAdrian Thurston2016-03-095-7/+55
| | | | | | | File names used in locations need to live beyond a program delete, but allocating them when exporting locations is too costly. Instead manage them in the colm program and allow them to be exported so they can live beyond the colm program. Will do something similar with the token text.
* clear any collect data from streamsAdrian Thurston2016-03-071-0/+5
|
* new stream objects can now collect text, accessible with ->tree (str type)Adrian Thurston2016-03-078-4/+84
| | | | | This is much faster than send to a line parser and then extracting the tree just to convert it to txt.
* compute correct position information in the top-level streamAdrian Thurston2016-03-071-6/+8
| | | | This will give us correct line information when parsing out of a string.
* prep code for managing stream name in the stream destructorAdrian Thurston2016-03-025-28/+30
| | | | | | | Requires strduping static names, and also copying names when stealing location information when a stream is pushed onto a stream. Can't yet delete (strings need after colm program is deleted). Need a solution for delated delete of a program's strings.
* plug a small string leakAdrian Thurston2016-03-011-0/+1
|
* added a fail-parsing flag that allows immediate abort from reduction actionsAdrian Thurston2016-03-012-0/+17
| | | | | This flag causes all alternate parsing paths to be aborted and pda_run to immedately fail. Tested with reduction actions.
* host adapters now have unique names are are accessed with func pointersAdrian Thurston2016-02-1911-50/+81
| | | | | This eliminates symbol collisions present when linking multiple colm programs into the same executable.
* bump version to 0.13.0.3colm-0.13.0.3Adrian Thurston2016-01-281-3/+3
|
* removed spurrious message introduced in prev commitAdrian Thurston2015-12-301-1/+1
|
* when pushing data in front of a stream, source the locationAdrian Thurston2015-12-261-2/+17
| | | | | When pushing data onto a stream in front of a source with locaiton information, grab that location information.
* added the $$ operator, which converts tree-to-str with attrsAdrian Thurston2015-12-157-8/+53
|
* clear the data trees as well as parse trees during reductionAdrian Thurston2015-12-023-6/+18
| | | | | This change brings memory consumption in line with a ragel+kelbt parse for large grammars and inputs.
* allow omission of location in reductionsAdrian Thurston2015-11-3011-42/+276
| | | | | | | There is code in here for omission of location and data in reductions. Unfortunately we cannot enable the omission of data unless we make changes to the backtracking. In the current implementation we push the token data with an mcopy from the data in the token.
* location fix for red actionAdrian Thurston2015-11-301-2/+1
|
* added an option, -b, for specifying the name of the program objectAdrian Thurston2015-11-275-6/+22
| | | | | This also triggers the omission of C adapter funcs, which need to be present only once in a binary that includes a colm code.
* reducer: cursor through rhs elements instead of starting from zeroAdrian Thurston2015-11-221-20/+90
|