float comparison functions with tolerance in ubertests

[sanikolaev]

Implemented functions to compare floats and strings containing floats with tolerance which is required for different CPU architectures, allowing for more flexible equality checks. Added a global flag to disable float tolerance for strict comparisons.

Master branch:

✘ snikolaev@dev2  ~/manticore_github/test   master  searchd -v
Manticore 14.5.0 41ecead3a@25111812 dev (columnar 8.1.0 e1522a2@25100213) (secondary 8.1.0 e1522a2@25100213) (knn 8.1.0 e1522a2@25100213) (embeddings 1.0.1)
Copyright (c) 2001-2016, Andrew Aksyonoff
Copyright (c) 2008-2016, Sphinx Technologies Inc (http://sphinxsearch.com)
Copyright (c) 2017-2025, Manticore Software LTD (https://manticoresearch.com)

 snikolaev@dev2  ~/manticore_github/test   master  php ubertest.php -i /usr/bin/indexer -s /usr/bin/searchd t 298
PERFORMING AUTOMATED TESTING

testing test_298, knn quantization... done; 1/1 subtests FAILED:
	subtest 1: query results mismatch; see test_298/report.txt

To re-run failed tests only:
php ubertest.php t 298

1 of 1 tests and 1 of 2 subtests failed, 0 tests skipped, 0.85 sec elapsed
THERE WERE FAILURES!

 snikolaev@dev2  ~/manticore_github/test   master  egrep "\-|+" test_298/report.txt|tail
 sphinxql-162> select id, knn_dist() from t where knn(emp,3,(-0.11468425393104553,0.3873162567615509,-0.26252424716949463,0.007096240296959877,0.24166066944599152,-0.24653761088848114,0.060873936861753464,0.23045268654823303));
 	id	knn_dist()
-	8	0.30891573
+	8	0.30891579
 	7	0.39149725
-	9	0.49754488
+	9	0.49754483
 3 rows in set

=== query 162 diff end ===

Branch ubertest/float-tolerance

 ✘ snikolaev@dev2  ~/manticore_github/test   ubertest/float-tolerance  php ubertest.php --float-tolerance -i /usr/bin/indexer -s /usr/bin/searchd t 298

PERFORMING AUTOMATED TESTING

testing test_298, knn quantization... done; 1/1 subtests OK

1 tests and 2 subtests succesful, 0 tests skipped, 0.92 sec elapsed
ALL OK

[glookka]

Test suite already has RoundFloatValues that is used for some cases, why not use it specifically for knn_dist case?

[sanikolaev]

Test suite already has RoundFloatValues that is used for some cases, why not use it specifically for knn_dist case?

Good idea. Done in 4460855

[glookka]

New version is better, however, it disables all float rounding using that new flag. This is wrong because RoundFloatValues must be used to round values in json result sets irregardless of the flag.

[sanikolaev]

@glookka

New version is better, however, it disables all float rounding using that new flag. This is wrong because RoundFloatValues must be used to round values in json result sets irregardless of the flag.

fixed in 222a5e2

[glookka]

As discussed, let's make this rounding mode disabled by default

[sanikolaev]

As discussed, let's make this rounding mode disabled by default

Done in
130871e