Only run semgrep for codemods with initial results

[drdavella]

Overview

Only run semgrep for codemods with initial results

Description

• This is an optimization that is intended to reduce runtime for the case where many/all codemods are requested
• The idea is to run semgrep once up-front to gather a list of applicable codemods and then to only run semgrep for those codemods that were identified in the initial pass
• There are further optimizations available here but this is a simple change intended to improve performance for this particular case
• Unfortunately this comes at the expense of doubling the semgrep cost of running a single codemod. We will address this later as it is not currently the bottleneck in performance.

[codecov]

Codecov Report

Merging #107 (deb18c6) into main (9b80e96) will increase coverage by 0.16%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #107      +/-   ##
==========================================
+ Coverage   95.71%   95.87%   +0.16%     
==========================================
  Files          62       62              
  Lines        2523     2524       +1     
==========================================
+ Hits         2415     2420       +5     
+ Misses        108      104       -4     

Files	Coverage Δ
src/codemodder/codemodder.py	96.80% <100.00%> (+0.18%)	⬆️
src/codemodder/sarifs.py	96.55% <100.00%> (+10.83%)	⬆️
src/codemodder/semgrep.py	94.73% <100.00%> (ø)

[andrecsilva]

I'm very skeptical this will have any actual impact on performance, otherwise it looks fine.

[drdavella]

@andrecsilva I gathered performance data that indicates that our runtime is completely dominated by invoking semgrep in certain environments. Anything we do to reduce the number of times we invoke semgrep is going to have a huge impact.