Files
@ 8b7c0ef62427
Branch filter:
Location: kallithea/docs/theme/nature/layout.html
8b7c0ef62427
1.0 KiB
text/html
search: make "repository:" condition work case-insensitively as expected
Before this revision, "repository:" condition at searching for "Commit
messages" never shows revisions in a repository, of which name uses
upper case letter.
Using ID for "repository" of CHGSETS_SCHEMA preserves case of
repository name at indexing. On the other hand, search condition
itself is forcibly lowered before parsing.
- files in repository "FOO" is indexed as "FOO" in "repository" field
- "repository:FOO" condition is treated as "repository:foo:
Then, indexing search itself is executed case-sensitively. Therefore,
"repository:FOO" condition never show revisions in repository "FOO".
But just making "repository" of CHGSETS_SCHEMA case-insensitive isn't
reasonable enough, because it breaks assumptions below, if there is
case-insensitive name collision between repositories, even though
Kallithea itself can manage such repositories at same time.
- combination of "raw_id" (= revision hash ID) and "repository" is
unique between all known revisions under Kallithea
CHGSETS_SCHEMA assumes this.
This unique-ness is required by Whoosh library to determine
whether index table should be updated or not for that repository.
- searching in a repository shows only revisions in that repository
Before this revision, this filtering is achieve by "repository:"
condition with case-preserved repository name from requested URL.
To make "repository:" search condition work case-insensitively as
expected (without any violation of assumptions above), this revision
does:
- make "repository" of CHGSETS_SCHEMA case-insensitive by
"analyzer=ICASEIDANALYZER"
- introduce "repository_rawname" into SCHEMA and CHGSETS_SCHEMA, to
ensure assumptions described above, by preserving case of
repository name
"repository_rawname" of SCHEMA uses not ID but TEXT, because the
former disable "positions" feature, which is required for
highlight-ing file content (see previous revision for detail).
This revision requires full re-building index tables, because indexing
schemas are changed.
Before this revision, "repository:" condition at searching for "Commit
messages" never shows revisions in a repository, of which name uses
upper case letter.
Using ID for "repository" of CHGSETS_SCHEMA preserves case of
repository name at indexing. On the other hand, search condition
itself is forcibly lowered before parsing.
- files in repository "FOO" is indexed as "FOO" in "repository" field
- "repository:FOO" condition is treated as "repository:foo:
Then, indexing search itself is executed case-sensitively. Therefore,
"repository:FOO" condition never show revisions in repository "FOO".
But just making "repository" of CHGSETS_SCHEMA case-insensitive isn't
reasonable enough, because it breaks assumptions below, if there is
case-insensitive name collision between repositories, even though
Kallithea itself can manage such repositories at same time.
- combination of "raw_id" (= revision hash ID) and "repository" is
unique between all known revisions under Kallithea
CHGSETS_SCHEMA assumes this.
This unique-ness is required by Whoosh library to determine
whether index table should be updated or not for that repository.
- searching in a repository shows only revisions in that repository
Before this revision, this filtering is achieve by "repository:"
condition with case-preserved repository name from requested URL.
To make "repository:" search condition work case-insensitively as
expected (without any violation of assumptions above), this revision
does:
- make "repository" of CHGSETS_SCHEMA case-insensitive by
"analyzer=ICASEIDANALYZER"
- introduce "repository_rawname" into SCHEMA and CHGSETS_SCHEMA, to
ensure assumptions described above, by preserving case of
repository name
"repository_rawname" of SCHEMA uses not ID but TEXT, because the
former disable "positions" feature, which is required for
highlight-ing file content (see previous revision for detail).
This revision requires full re-building index tables, because indexing
schemas are changed.