Unconditionally add testCaseId to speed up code coverage massively #783
Conversation
Codecov Report
@@ Coverage Diff @@
## master #783 +/- ##
============================================
- Coverage 84.29% 84.28% -0.01%
+ Complexity 1074 1072 -2
============================================
Files 56 56
Lines 3234 3233 -1
============================================
- Hits 2726 2725 -1
Misses 508 508
Continue to review full report at Codecov.
|
| if (empty($this->lineCoverage[$file][$k]) || !in_array($testCaseId, $this->lineCoverage[$file][$k], true)) { | ||
| $this->lineCoverage[$file][$k][] = $testCaseId; | ||
| } | ||
| $this->lineCoverage[$file][$k][] = $testCaseId; |
There was a problem hiding this comment.
If this is safe, then the checks on 83/91 (pre-patch) 81/89 (post-patch) I think should be safe to remove as well, the principle behind adding those was cargo-culting over from this one
There was a problem hiding this comment.
Not sure what happens in this case but it's possible for the same test to be added to this array twice now.
If the array was keyed by the test id, you could set the key to be the id and the value to be true ? Then use array_keys before processing.
Not checked if adding the same test matters though
With this change, both PHP Code Coverage's and PHPUnit's test suite still passes.
It makes the PHPUnit test suite run in approximately ⅔rds of the time, with no differences in output:
PHPUnit's test suite (
$ php -dxdebug.mode=coverage,profile -dxdebug.start_with_request=yes -dhtml_errors=0 -dmemory_limit=2G ./phpunit --testsuite=unit --coverage-html=/tmp/cov-after.html):Time: 00:52.085, Memory: 82.00 MBTime: 00:33.547, Memory: 82.00 MB, a reduction of 36%.The valgrind/cachegrind profile shows it is using only 27% of its previous instructions.