Improve performance

[dineshba]

Co-authored-by: @aswinkarthik

We introduced go routines in pattern detector and file content detector to
make things in parallel.

Results:

For a codebase which had 600 newly added files

Before our change:

time ~/.talisman/bin/talisman_darwin_amd64 --githook pre-commit >/dev/null

real    0m14.029s
user    0m12.035s
sys     0m2.216s

After our change:

time ./talisman_darwin_amd64 --githook pre-commit >/dev/null
real    0m7.797s
user    0m19.857s
sys     0m2.336s

Method level metrics:

Pattern Detector:
    Before: 5.498711305s
    After: 1.898453802s

Content Detector:
    Before: 639.769275m
    After: 221.733402ms

[svishwanath-tw]

Lot of repeated code here.
I wonder if repetition could be avoided ?

[dineshba]

Line 96 to 112, we are just creating 3 structs and sending to channel. I am not very sure how to reduce that. If you have any idea, please let me know.

[dineshba]

Refactored this. Please review it

[dineshba]

This will help to close this #152
Thanks @svishwanath-tw for merging

[karuppiah7890]

I haven't checked the PR completely. But from an overview, I'm just curious about tests for this PR. Tests for the happy path and to show that performance has increased. Like, how do I say if my performance has increased by this PR, and how do I know if it gets decreased by some other code later? I haven't done such testing before, just curious. Some ideas - probably do some profiling? Or use some other parameter that relates to performance and check it?

cc @svishwanath-tw @dineshba @aswinkarthik

[svishwanath-tw]

@karuppiah7890 are you saying we need to add tests to the test suite ?

[karuppiah7890]

If it's possible, yes. I mean, I would need some measure to say the performance has increased, it can be in any form. Numbers always speak for themselves. Or else we will never know how much performant the code is compared to before. And such a metric would be useful in the future too

[svishwanath-tw]

Does this help at all ? https://stackoverflow.com/questions/13003749/number-of-elements-in-a-channel

[svishwanath-tw]

https://tour.golang.org/concurrency/4
According to this article: for ... range will also work in the scenario.

[dineshba]

https://tour.golang.org/concurrency/4
According to this article: for ... range will also work in the scenario.

We are reading from two channels simultaneously. So we have to use select. Refer https://stackoverflow.com/questions/20593126/reading-from-multiple-channels-simultaneously-in-golang

[svishwanath-tw]

@dineshba : The two processing methods seem to have the same signature.
Why not extend the contentType abstraction to process ignored content too and just have one processor ?
The advantage is that you can get away with 1 channel right ?