Is there a way to hook in to a build target and run some fun

Question

Is there a way to hook in to a build target and run some functionality before afterwards

witty-crayon-22786 · Answer

before is possible by adding a dependency on an `experimental shell command` <https www pantsbuild org docs run shell commands> but note that the process will run in a sandbox

witty-crayon-22786 · Answer

after currently isn t possible

green-match-66337 · Answer

hmmm trying to restructure after lambda target build as PEX seems to add 400ms cold start based on the structure

witty-crayon-22786 · Answer

ah sorry i assumed a particular usecase testing you can definitely use an `experimental shell command` to consume a `pex binary` and manipulate it

witty-crayon-22786 · Answer

so something like ```pex binary name= a experimental shell command name= b dependencies= a archive name= c dependencies= b ``` would allow you to create an archive after manipulating a PEX but feeding a PEX through a script like this loses type information it s no longer a PEX from Pants perspective

witty-crayon-22786 · Answer

i m not familiar with whether lambda keeps environments around but in the `venv` <https www pantsbuild org docs reference pex binary codeexecution modecode|execution mode> latency should be within 50 ms of an existing venv after the first run

green-match-66337 · Answer

green-match-66337 · Answer

This is what I m thinking of following

green-match-66337 · Answer

Seems the package size in lambda doesnt matter as much as how accessible the packages are

witty-crayon-22786 · Answer

`experimental shell command` should be roughly equivalent to `genrule`

witty-crayon-22786 · Answer

but if there is an alternative layout of a PEX that ends up being significantly faster reporting an issue on <https github com pantsbuild lambdex > would be appreciated i m not aware of why the alternate layout would be faster the blog post mentions a different please specific reason for doing this

green-match-66337 · Answer

Hard to tell but from what I have seen AWS use some underlying caching mechanism which is why the docker image size doesnt increase cold starts too much for a docker image lambda function I imagine its the same for Python so the size matters less and they likely optimize for SAM

happy-kitchen-89482 · Answer

Is that 400ms when compared to a hand constructed lambda zip containing all the same requirements etc

green-match-66337 · Answer

Correct hand constructed being SAM

happy-kitchen-89482 · Answer

Hmm probably we should change our implementation to construct lambdas the same way SAM does using their CLI even rather than with lambdex