Non-executable shell script with shebang line should be detected as "shell"

[l0b0]

Actual behaviour:

$ lorri init
[…]
$ identify-cli .envrc 
["file", "non-executable", "text"]
$ chmod u+x .envrc
$ identify-cli .envrc 
["bash", "executable", "file", "shell", "text"]

Expected behaviour: the results should be identical for the executable and non-executable versions of the file, because the first line in the file is a valid shell script shebang line: #!/usr/bin/env bash.

[asottile]

the README is short and addresses this

[l0b0]

@asottile I'm not sure I understand. The relevant example uses an executable file:

>>> identify.tags_from_path('/path/to/file-with-shebang')
{'file', 'text', 'shell', 'bash', 'executable'}

So this is an issue with the integration with pre-commit?

[asottile]

nope, this is a "you stopped reading before you got to the relevant part" problem:

If executable, the shebang is read and the interpreter interpreted

[l0b0]

OK, so it's not working as expected. Why does it not read the shebang line of non-executables?

[l0b0]

Tools like IDEs and ShellCheck do read the shebang line of non-executables, after all. So does the file command:

$ file .envrc 
.envrc: Bourne-Again shell script, ASCII text executable

[asottile]

the point of a shebang is to indicate to the OS how to execute it -- it is only relevant when it is executable

parsing shebangs of non-executables would also mean everything has to read every file always which is slow

this is not up for debate so please do not waste your time and my time