replace own wildcard matcher with QRegularExpression combining all wildcards
In addition add validator to index tool, only ? and * wildcard stuff is valid fix some syntax files that had e.g. , instead of ; and spaces mixed in remove asterisk definition, that will never match as it tried to match the file path and is anyways 'very' basic add auto-test for new regex stuff
make && make test still works, autotest added
| No Linters Available |
| No Unit Test Coverage |
| Buildable 2244 | |
| Build 2262: arc lint + arc unit |
Rational for the static convertExtensionsToRegularExpression method: Allows to nice unit tests + allows e.g. KTextEditor to use that later, too, as replacement for its own copy of the wildcard matcher code for the modelines.
all code is now MIT
only non-MIT parts are e.g. some generator scripts for highlighting and many highlighting data files
improve definitionForName code, already for matching take into account the priority
that saves some matching time and avoid later sorting
The old wildcard matcher existed for performance reasons. Did you check?
PS: When doing review requests, could you please have a nice subject, and not put very long lines into its stead? :-p Especially since this line will also appear in Davids release notes?
Some trivial QBenchmark stuff tells
QBENCHMARK {
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.c"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.sdfdsf"));
m_repo->definitionForFileName(QLatin1String("ksldjfklsdjfkl.ada"));
m_repo->definitionForFileName(QLatin1String("12323.cklsjdf"));
}tells:
new: between 4 and 5 msecs per iteration
old: between 5 and 6 msecs per iteration
I think the main issue in the "very old code" was, that we did a QRegExp per extension, not a combined one.
For the accessor: I actually would prefer a const QRegularExpression & as return value, but most other methods use values to hide the internals.
I will add the internal stuff like the wildcard check and the removed/fixed hls.
I will not add the regex changes and any public API, I want to think about that once more first.
That isn't that urgent anyways. KTextEditor uses its own code for that and the only consumer is the definitionForName ATM.
I fixed the hl files, I removed the stray Asterisk stuff, that was no proper HL for such files, they are more or less only normal .ini files, added .conf as end for ini.
Indexer is checked in, too.
I will do a new request for a properly redone matching for post 5.50.
Asterisk removal:
https://commits.kde.org/syntax-highlighting/8cc16753881d904c03f32c444f503d186d02343a
Indexer improvements:
https://commits.kde.org/syntax-highlighting/3d1b8dc72191de03a97f6c2a74cc5323dfff7b94