CurlyMoo / rules / 41

97%

push

travis-ci-com

<a href="https://github.com/CurlyMoo/rules/commit/<a class=hub.com/CurlyMoo/rules/commit/<a class="double-link" href="https://git"><a class=hub.com/CurlyMoo/rules/commit/7863ea54e7739f62514819dfdb530a07bc23e195">7863ea54e<a href="https://github.com/CurlyMoo/rules/commit/7863ea54e7739f62514819dfdb530a07bc23e195"><a href="https://github.com/CurlyMoo/rules/commit/7863ea54e7739f62514819dfdb530a07bc23e195">&quot;&gt;Kinda revert &lt;/a&gt;&lt;a class=&quot;double-link&quot; href=&quot;https://github.com/CurlyMoo/rules/commit/&lt;a class=&quot;double-link&quot; href=&quot;https://github.com/CurlyMoo/rules/commit/4db5a4befd6156cb57e127f5f2c313176220fc87&quot;&gt;4db5a4bef</a>">4db5a4bef</a><a href="https://github.com/CurlyMoo/rules/commit/7863ea54e7739f62514819dfdb530a07bc23e195"> to prevent memory fragmentation

With commit </a><a class="double-link" href="https://github.com/CurlyMoo/rules/commit/<a class="double-link" href="https://github.com/CurlyMoo/rules/commit/759f0350575ee606f36e56705add13c008ad43e7">759f03505</a>">759f03505</a><a href="https://github.com/CurlyMoo/rules/commit/7863ea54e7739f62514819dfdb530a07bc23e195">
the AST size can be calculated up front. That means
there is no gain in eating the original rule. Instead,
eating the original rule offers new issues, which is
memory fragmentation.

Eating the original rule can mean that a new block
of memory is allocated according to the newly prefered
size, next to the original memory block. So instead of
using less memory this may shorty use more memory when
reallocating to a smaller size.

Another downside of constantly reallocating memory is
that the parsing speed can be up to 30 times slower.
In the old situation the same rule was parsed in 5
seconds. Reverting the eating logics increases that
speed back to 0.17 seconds.

The last improvement with this commit is that the
original preperation doesn&#39;t require a new memory
allocation. Instead, the original rule is overwritten
with new preperation logic.

So,
- One allocation for the ruleset read from the filesystem.
- One allocation for the AST.
- One reallocation when a rule block has been parsed.
- One free when all rules are parsed.

The best approach to keep memory allocation low is to
split the ruleset in multiple files of let's say 2048Kb

2388 of 2441 relevant lines covered (97.83%)

1151.1 hits per line