| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
| |
| |
| |
| | |
This is a Tracker SPARQL extension present in the previous parser.
|
| |
| |
| |
| | |
This is a Tracker extension to SPARQL.
|
| |
| |
| |
| | |
This is a tracker extension to SPARQL.
|
| |
| |
| |
| | |
This is syntax traditionally allowed by Tracker.
|
| |
| |
| |
| | |
This is a Tracker extension to the SPARQL1.1 syntax.
|
| | |
|
| | |
|
| | |
|
| | |
|
| |
| |
| |
| |
| |
| | |
For sequential property paths it will be more convenient to have the
ObjectList node available before processing the property path, so we
can explode those properly into intermediate blank nodes.
|
| |
| |
| |
| | |
Use the equivalent EXCEPT sqlite syntax.
|
| |
| |
| |
| | |
If glib >= 2.51.0 is available, we can implement this.
|
| |
| |
| |
| | |
This is a syntax extension we used to accept, so bring it back.
|
| |
| |
| |
| |
| | |
Tracker used to accept SubSelect in BrackettedExpression, bring
that back.
|
| |
| |
| |
| |
| | |
Tracker used to accept plain Expression in SelectClause, i.e.
no parentheses, and "AS ?var" being optional. Bring that back.
|
| |
| |
| |
| | |
So the sparql parser can defer optional arguments to it.
|
| | |
|
| |
| |
| |
| |
| |
| |
| |
| | |
This is somewhat similar to GString, except it allows for adding
extension points at random places that allow for inserting stuff
mid-string without relocating the whole thing.
It is expected that SQL query construction will use this facility.
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This struct takes the query text and a root rule that will work as the
starting point to generate the expression tree that represents the query.
This expression tree has a virtually identical structure to the formed
by the rules themselves, with the only difference that ?/+/* rules may
have none or many nodes representing a single rule.
The parser does not do anything beyond a syntactic interpretation of the
query, the expression tree nodes contain a pointer to the rule that
generated them, plus start/end points of the text in the query, so an
upper layer can do the actual ontology validations, and generate the SQL.
The two entry points are tracker_sparql_parse_query/update, that correspond
to the QueryUnit/UpdateUnit entry points defined in the SPARQL grammar.
|
|/
|
|
|
|
| |
The grammar rules are based on
https://www.w3.org/TR/2013/REC-sparql11-query-20130321/#sparqlGrammar
and reflect all of the Sparql 1.1 syntax, supported by Tracker or not.
|
|\ |
|
| |
| |
| |
| |
| |
| |
| | |
The previous change did not leave the expected .0 symlinks.
'soversion' is the actual version linked against. To match the scheme
used by libtool, we need to give the libraries a 'version' as well.
|
|/
|
|
|
|
|
| |
Export those in tracker-sparql.pc, so users may find out the install
details.
Related: https://gitlab.gnome.org/GNOME/tracker-miners/issues/19
|
|\ |
|
| |
| |
| |
| |
| | |
Due to a mistake in the test runner, it was possible for tests to
fail but appear to have succeeded.
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This warning is disabled by default at warning level 1.
Multiple distros are building Tracker with -Werror=format-security and
discovering regressions, we have had reports from Debian, Fedora and
NixOS so far, so let's enable it by default.
|
| | |
| | |
| | |
| | |
| | | |
If the data manager errors out, trying to free the connection afterwards
will trigger warnings trying to shut it down.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Those may happen in certain circumstances on monitored dirs, but are safe
to simply ignore. This leaves the warning for EVENT_MOVED which can only
be a bug in GLib since we use G_FILE_MONITOR_WATCH_MOVES.
Closes: https://gitlab.gnome.org/GNOME/tracker/issues/54
|
| | |
| | |
| | |
| | |
| | | |
Fixes build error with -Dfts=false, and a warning that must have
been there forever.
|
| | |
| | |
| | |
| | |
| | |
| | | |
This function is no longer used, so makes no sense to test for
it. This test used to break on CI with the default docker images
because of user permissions, so this obstacle is removed.
|
| | |
| | |
| | |
| | |
| | | |
We haven't used configuration keyfiles for a long long time, just
stop trying to delete them.
|
|\ \ \ |
|
|/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Sqlite >= 3.25.0 got stricter 'alter table ... rename ...' behavior
where the renaming affects all references to the table across the
database. This is at odds with our fts_view view on schema updates
due to ontology changes, as we first migrate (rename current table,
create new one, insert previous content, drop old table) all resource
tables before doing the FTS table/view updates. This causes rename
failures due to the fts_view referencing the renamed and dropped
tables.
Change the ontology update code so we delete the FTS table/view
before changing resource tables in case of FTS updates. This makes
the behavior correct both before and after the Sqlite change.
https://gitlab.gnome.org/GNOME/tracker/issues/40
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Both libtracker-fts and libtracker-data end up requiring includes
from tracker-sparql.h, but we have to generate tracker-generated.h
first.
It's pretty terrible to add intermediate targets as a dependency
outside the libtracker-sparql directory, but it doesn't seem we
can do better.
Closes: https://gitlab.gnome.org/GNOME/tracker/issues/52
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This reverts commit 20b9f3673d16d07070619cb0ff770d0dabbcc74c.
Gnome infrastructures should already have meson >= 0.48.0, let's give it
another try.
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Instead of explicitly linking to intermediate libraries in
src/libtracker-sparql, have it depend on the dependency declared there.
Should provide the right -L/-l arguments when using the
libtracker_sparql_dep dependence.
Failure seen in:
https://gitlab.gnome.org/GNOME/gnome-build-meta/-/jobs/125673
|
| | |
| | |
| | |
| | |
| | |
| | | |
The scheme adopted for development releases is "alphaN" for unstable
releases, and "betaN" for release candidates. major/minor/micro version
stay fixed in the targeted version.
|
| | |
| | |
| | |
| | |
| | |
| | | |
Leaving to '0' is not enough if several versions of the library are
installed (say, in different prefixes). Add back the library soversion,
similarly to how it was done in autotools.
|
| | |
| | |
| | |
| | |
| | |
| | | |
We generate that file ourselves, so configure_file() is unneeded as the
file requires no substitutions. Fixes a warning on recent meson, as the
configuration should not be empty.
|
| | |
| | |
| | |
| | |
| | |
| | | |
Just had to look a little bit above...
This reverts commit f7e4ea51a9a62e0cddac1a9407b5a6f9a91545f4.
|
| | |
| | |
| | |
| | | |
This isn't even used, but that's no reason to leave it out of docs.
|
| | |
| | |
| | |
| | |
| | | |
There were also some references to no longer existing ontology, that's
been wiped out.
|
| | |
| | |
| | |
| | |
| | | |
That is as much as we say about them, gets rid of warnings during docs
generation.
|
| | |
| | |
| | |
| | | |
The build_always option is deprecated since meson 0.38.0
|
| | |
| | |
| | |
| | | |
So we get proper links to glib types.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The docs were not going through gtkdoc-scangobj, and the libtracker-sparql
docs were just looking in source dir while it should also look for gtk-doc
comments in generated files from vala.
Now that we're there, use include_directories() to get rid of relative
paths.
|
| | |
| | |
| | |
| | |
| | | |
The old unorm_normalize() function has been deprecated. This replacement
function is available on ICU >= 4.4, so we should be fine dependency-wise.
|
| | |
| | |
| | |
| | | |
Modern vala recommends this.
|