.optimize() no longer cleans up _docsize

This isn't necessary any more since the new .rebuild_fts()
method can achieve the same thing.

Refs #155, #153
This commit is contained in:
Simon Willison 2020-09-08 15:18:12 -07:00
commit 64799df78b
4 changed files with 22 additions and 38 deletions

View file

@ -1228,6 +1228,10 @@ This method can be called on a table that has been configured for full-text sear
db["dogs_fts"].rebuild_fts()
This runs the following SQL::
INSERT INTO dogs_fts (dogs_fts) VALUES ("rebuild");
Optimizing a full-text search table
===================================
@ -1240,10 +1244,6 @@ Once you have populated a FTS table you can optimize it to dramatically reduce i
This runs the following SQL::
INSERT INTO dogs_fts (dogs_fts) VALUES ("optimize");
DELETE FROM [dogs_fts_docsize] WHERE id NOT IN (
SELECT rowid FROM [dogs_fts]);
That ``DELETE`` statement cleans up rows that may have been created by `an obscure bug <https://github.com/simonw/sqlite-utils/issues/153>`__ in previous versions of ``sqlite-utils``.
Creating indexes
================

View file

@ -951,20 +951,6 @@ class Table(Queryable):
table=fts_table
)
)
self.db.execute(
textwrap.dedent(
"""
DELETE FROM [{table}_docsize] WHERE {column} NOT IN (
SELECT rowid FROM [{table}]);
"""
)
.strip()
.format(
# FTS5 uses 'id' but FTS4 uses 'docid'
column=self.db["{}_docsize".format(fts_table)].columns[0].name,
table=fts_table,
)
)
return self
def search(self, q):

View file

@ -1,4 +1,5 @@
import pytest
from sqlite_utils import Database
from sqlite_utils.utils import sqlite3
search_records = [
@ -199,3 +200,20 @@ def test_rebuild_fts_invalid(fresh_db, invalid_table):
# Raise OperationalError on invalid table
with pytest.raises(sqlite3.OperationalError):
fresh_db[invalid_table].rebuild_fts()
@pytest.mark.parametrize("fts_version", ["FTS4", "FTS5"])
def test_rebuild_removes_junk_docsize_rows(tmpdir, fts_version):
# Recreating https://github.com/simonw/sqlite-utils/issues/149
path = tmpdir / "test.db"
db = Database(str(path), recursive_triggers=False)
licenses = [{"key": "apache2", "name": "Apache 2"}, {"key": "bsd", "name": "BSD"}]
db["licenses"].insert_all(licenses, pk="key", replace=True)
db["licenses"].enable_fts(["name"], create_triggers=True, fts_version=fts_version)
assert db["licenses_fts_docsize"].count == 2
# Bug: insert with replace increases the number of rows in _docsize:
db["licenses"].insert_all(licenses, pk="key", replace=True)
assert db["licenses_fts_docsize"].count == 4
# rebuild should fix this:
db["licenses_fts"].rebuild_fts()
assert db["licenses_fts_docsize"].count == 2

View file

@ -1,20 +0,0 @@
import pytest
from sqlite_utils import Database
import sqlite3
@pytest.mark.parametrize("fts_version", ["FTS4", "FTS5"])
def test_optimizes_removes_junk_docsize_rows(tmpdir, fts_version):
# Recreating https://github.com/simonw/sqlite-utils/issues/149
path = tmpdir / "test.db"
db = Database(str(path), recursive_triggers=False)
licenses = [{"key": "apache2", "name": "Apache 2"}, {"key": "bsd", "name": "BSD"}]
db["licenses"].insert_all(licenses, pk="key", replace=True)
db["licenses"].enable_fts(["name"], create_triggers=True, fts_version=fts_version)
assert db["licenses_fts_docsize"].count == 2
# Bug: insert with replace increases the number of rows in _docsize:
db["licenses"].insert_all(licenses, pk="key", replace=True)
assert db["licenses_fts_docsize"].count == 4
# Optimize should fix this:
db["licenses_fts"].optimize()
assert db["licenses_fts_docsize"].count == 2