AnchoredSetVariable::resolve is called for every rule
(see RuleWithOperator::evaluate). The previous implementation allocated
a new copy of every variable, which quickly added up. In my tests,
AnchoredSetVariable::resolve function consumed 7.8% of run time.
AnchoredSetVariable (which is a multimap) values are never changed,
only added. This means it's safe to store them in std::shared_ptr,
and make resolve return shared_ptr pointing to the same object.
Other resolve implementation could also use this optimization by not
allocating new objects, however, they are not hot spots, so this
optimization was not implemented there.
In my benchmark, this raises performance from 117 requests per second to
131 RPS, and overhead is lowered from 7.8% to 2.4%.
As a bonus, replacing plain pointer with smart pointers make code
cleaner, since using smart pointers makes manual deletes no longer necessary.
Additionally, VariableOrigin is now stored in plain std::vector,
since it's wasteful to store structure containing just two integer
values using std::list<std::unique_ptr<T>>.
This issue was initially reported by @michaelgranzow-avi on #2296.
@airween made an initial attempt to provide a fixed at #2107; As a
consequence of the pull request review - provided by @victorhora,
@zimmerle, and @michaelgranzow-avi - @airween made a second attempt
at #2297. After reviewing by @martinhsv, @zimmerle, I have absorbed
the essential pieces from @airween patch into this one.
This patch differs from @airween's because @airween's patches were
partially working: Key exclusions with regex weren't covered, same
for anchored variables (e.g. ARGS). During the review, I have
highlighted the importance of having elementary test cases. A simple
test case on ARGS could spot the issue. Since that is an important
fix, I don't want to hold this for one more review cycle; therefore,
I am committing the fix myself.
Thank you all involved in the solution of this very own issue.
This commit fixes quite a few odd things in regex code:
* Lack of encapsulation.
* Non-method functions for matching without retrieving all groups.
* Regex class being copyable without proper copy-constructor (potential UAF
and double free due to pointer members m_pc and m_pce).
* Redundant SMatch::m_length, which always equals to match.size() anyway.
* Weird SMatch::size_ member which is initialized only by one of the three matching
functions, and equals to the return value of that function anyways.
* Several places in code having std::string value instead of reference.
