Improve importRecursive performance with a regexp

[julienduchesne]

Instead of parsing the whole AST trying to find imports, find them with a regexp. This is significantly faster for very nested projects
I added a benchmark in the tests. Here's the result:

$ benchstat bench1.txt bench2.txt
name                 old time/op  new time/op  delta
TransitiveImports-8  0.00ns ±47%  0.00ns ±30%  -25.85%  (p=0.000 n=92+90)

In real-life cases, I have gotten 10x faster run time (2.5s to 200ms) on some of our heavier projects (which are the ideal cases for caching, where this function is used)

[sh0rez]

without having looked into too much detail here, wouldn't this be vulnerable to false positives?

bad: 'import "foobarbaz"'

[julienduchesne]

without having looked into too much detail here, wouldn't this be vulnerable to false positives?

bad: 'import "foobarbaz"'

definitely, but there two uses for this function:

• TransitiveImports: This function will fail if any of the imports are not found. You could have a commented out import that does resolve though, this could indeed be annoying
• getSnippetHash: Used for environment caching. For this one, the only important thing is to get a consistent hash that changes consistently when any of the imports changes. In that case, false positives aren't a big deal but speed is

I'm wondering if I should keep both strategies. Using the AST for TransitiveImports and a regexp for calculating the hash of an env. What do you think?

[julienduchesne]

Converting to a draft. Will work on a different approach

[julienduchesne]

Implemented in #759