Delete the source code from the token cache #51
Conversation
…ng statistics. This greatly reduces memory usage when running coverage against many files. Paired with Mark French.
| @@ -169,6 +169,8 @@ class PHP_CodeCoverage_Report_HTML_Node_File extends PHP_CodeCoverage_Report_HTM | |||
| ); | |||
|
|
|||
| $this->calculateStatistics(); | |||
|
|
|||
| PHP_Token_Stream_CachingFactory::delete($path); | |||
There was a problem hiding this comment.
In most situations both the Clover XML and HTML report are generated. Both report generators leverage the token cache.
|
So this becomes a balance between memory usage and disk IO. I'm running into serious memory usage issues right now trying to run code coverage against my entire code base. php-code-coverage uses over 1GB of memory then hits the memory limit and dies. I'm willing to take the extra hit of a file read to significantly reduce memory usage. I haven't noticed any difference in code coverage generation time from this change. |
|
That's true, but it requires that the tokenized representation of ALL source files be kept in memory for the entire duration of BOTH report generation runs. This delete allows only a single file to be kept in working memory at a time, which prevents the current behavior of steadily growing (read: runaway) memory consumption. |
|
Yes, it is a tradeoff between memory and CPU/disk. That is why the caching can be turned off for PHP_CodeCoverage. |
|
That's great to hear caching can be disabled. I looked through the code and don't see anything that would make caching happen conditionally. I must be missing something... how do I turn it off? Thanks. |
|
PHPUnit 3.6 introduces a |
Delete the source code from the token cache when we're done calculating statistics. This greatly reduces memory usage when running coverage against many files. When running tests against a subset of my codebase, this change reduced memory usage from ~900MB to ~300MB. I paired on this work with my coworker, Mark French.