Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

had some fun implementing python run in v2: <https://github.com/pantsbuild/pants/pull/8236>

Yay! It looks like this implementation can be used to implement python binary? If I’m reading correctly, python run == python binary + actually running that PEX

We’ll probably want to extract out a lot of the duplicated logic with python_test_runner.py, such as the source root stripping

that's correct, this was hacked together in a few hours to show off to a twitter engineer

removing the duplicated logic should also reduce the diff size considerably

right now i'm focusing on python run because python binary requires creating a file in dist/ and i don't want to figure that out just yet

what's the rust way to turn a `Vec&lt;Future&lt;_,_&gt;&gt;` into a `Future&lt;Vec&lt;_&gt;,_&gt;`?

and are we doing something like this anywhere in the enigne alrady?

almost definitely somewhere in the remote execution backend, i’ll look

If I want to get a Snapshot (and eventually FileContents) for a file for which I have a relative path from the buildroot, do I use `PathGlobs`, with the understanding that a single fixed path is a degenerate special case of a glob? Or is there a `Path` intrinsic?

I can't figure out from the code, for example, how build file parsing handles something like `sources=['path/to/source1.java', 'path/to/source2.java']`

it’s the first one, where a single path is a “degenerate” case of a glob, although i disagree on calling it degenerate

if the file denoted by the path doesn’t exist, then it resolves to a set of 0 files in the same way that a set of globs matching nothing would 

that’s not necessarily the only way to think about it but that’s what goes through my head 

I see, it's a bit weird to access something called `directory_digest` on a Snapshot when I know it's just a single file

yes, i absolutely agree and i think not having simpler abstractions for these pretty-consistent filesystem concepts has hampered v2 rule development 

I guess that's a misnomer in general since a Snapshot can represent any collection of files

a directory digest can also represent any collection of files 

I mean a globs(".py") is not a directory, it's a collection of files in a directory

it represents "some set of 0, 1 or more files"

it represents the _contents_ of a directory 

does it really not have directories in there? i thought snapshots had directory paths too? wow

They do `class Snapshot(datatype([('directory_digest', Digest), ('files', tuple), ('dirs', tuple)]))`

But isn't the digest the digest of all the files and dirs covered by the snapshot?

Right, they do not necessarily represent the contents of a directory

For example, we go from `PathGlobs` to `Snapshot`

i was thinking that they represent the contents of a virtual directory 