The (experimental) inverted index writes/reads files different from the
standard files written by the other skip indexes. The original problem
was that with database engine "ordinary", DROP TABLE of a table with
inverted index finds unknown files in persistence and complains. The
same will happen with engine "atomic" but deferred. As a hotfix, the
error was silenced by explicitly adding the four files created in a
specific test to the deletion code.
This PR tries a cleaner solution where all needed files are provided via
the normal checksum structure. One drawback remains which is that the
affected files were written earlier and we don't have their checksums
available. Therefore, the inverted index is currently excluded from
CHECK TABLE.
Minimal repro:
SET allow_experimental_inverted_index = 1;
DROP TABLE IF EXISTS tab;
CREATE TABLE tab(s String, INDEX af(s) TYPE inverted(2)) ENGINE = MergeTree() ORDER BY s;
INSERT INTO tab VALUES ('Alick a01');
CHECK TABLE tab;
DROP TABLE IF EXISTS tab;
run ./clickhouse-test with --db-engine Ordinary
* save format string for NetException
* format exceptions
* format exceptions 2
* format exceptions 3
* format exceptions 4
* format exceptions 5
* format exceptions 6
* fix
* format exceptions 7
* format exceptions 8
* Update MergeTreeIndexGin.cpp
* Update AggregateFunctionMap.cpp
* Update AggregateFunctionMap.cpp
* fix
TODO (suggested by Nikolai)
1. Build query plan fro current query (inside storage::read) up to WithMergableState
2. Check, that plan is simple enough: Aggregating - Expression - Filter - ReadFromStorage (or simplier)
3. Check, that filter is the same as filter in projection, and also expression calculates the same aggregation keys as in projection
4. Return WithMergableState if projection applies
3 will be easier to do with ActionsDAG, cause it sees all functions, and dependencies are direct (but it is possible with ExpressionActions also)
Also need to figure out how prewhere works for projections, and
row_filter_policies.
wip
with the explicit data loading and verification.
If the function fails, the exception will re-throw upper,
cancelling the fetch in the handling of the replicated log entry,
but this event will also wake the PartCheckThread, which will
issue a re-fetch.
- Updated the docs to make everything clear.
- Multiple small logger fixes.
- Changed the attach_part command -- now it's after check for the
covering parts -- motivation is to do less work with the checksums
fetching.
- Better logging in the integration test.
