Update replace-placeholder.sh for much faster startup

[dvusboy]

Speed up the replacement loop

• only look at files with the "$FROM" pattern

[dvusboy]

A quick test with v4.5.3 shows these results:

Current:

root@4f3f81c57237:/calcom/apps# FROM='http://localhost:3000'; TO='https://cal.example.com'; time find web/.next/ web/public/ -type f | while read file; do sed -i "s|$FROM|$TO|g" "$file"; done

real	1m22.762s
user	0m52.981s
sys	0m44.544s

New:

root@a6f1f3600029:/calcom/apps# FROM='http://localhost:3000'; TO='https://cal.example.com'; time for file in $(egrep -r -l "$FROM" web/.next/ web/public/); do sed -i -e "s|$FROM|$TO|g" "$file"; done

real	0m3.356s
user	0m1.487s
sys	0m1.958s

[codablock]

Optimization generally LGTM, but without knowing this project well enough, I'd be worried how you got to so many files in these directories that a simple looped sed ends up needing 1m22 to process.

@dvusboy Did you observe these times on custom builds only? Did you maybe have some kind of leftovers from older builds locally that got copied into the final image?

[dvusboy]

Well, 2 points:

• To get the timing difference, I just run the script in-place, changed and not changed. It's pretty easy to do by overriding CMD to the Docker image with /bin/bash with docker run.
• It is pretty obvious that the changes would speed things up by reducing the number of files needing changes. The extent was a little surprising.

I didn't change much to the sed line, just explicitly prefixing the script with -e.

[codablock]

Ah ok, so you ran it outside of the Docker build? That would explain the huge amount of time needed initially (I assume a local build ends up having much more/larger files then a Docker build).

As said, the optimisation itself is a LGTM.