Q: how to pass a string as stdin to Runners' run()?

[adswa]

The docstring of the Runner's run method state the following about the stdin= parameter:

datalad/datalad/runner/runner.py

Lines 91 to 98 in 66a8b77

	stdin : file-like, string, bytes, Queue, or None
	If stdin is a file-like, it will be directly used as stdin for the
	subprocess. The caller is responsible for writing to it and closing it.
	If stdin is a string or bytes, those will be fed to stdin of the
	subprocess. If all data is written, stdin will be closed.
	If stdin is a Queue, all elements (bytes) put into the Queue will
	be passed to stdin until None is read from the queue. If None is read,
	stdin of the subprocess is closed.

However, if I provide a string to it (I tried toying around in the context of #6514), I get

Traceback (most recent call last):
  File "/usr/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
    self.run()
  File "/home/adina/repos/datalad/datalad/runner/runnerthreads.py", line 172, in run
    succeeded = self.write(data)
  File "/home/adina/repos/datalad/datalad/runner/runnerthreads.py", line 246, in write
    written += os.write(
TypeError: a bytes-like object is required, not 'str'

I feel like I might be missing something fairly obvious. How can I provide a string as stdin to a runner.run() call?

[mih]

It seems the documentation is wrong, it must be bytes. I can only find usage of this feature where bytes are used specifically.

"somestr".encode('utf-8')

[bpoldrack]

ATM, I consider this a bug. Runner explicitly checks whether stdin is str or bytes and proceeds to put it in the queue either way. I guess that's where it should be encoded in case of str.

[christian-monch]

Maybe we should change the description and the check and only allow bytes-input. That would force the user to explicitly encode the string (instead of us encoding it with a fixed encoding, or the system-preferred encoding, or the encoding from the protocol keyword arguments)

[bpoldrack]

instead of us encoding it with a fixed encoding, or the system-preferred encoding, or ...

Seems like a good point. Will change #7155 accordingly.

[yarikoptic-gitmate]

Issue fixed in 0.17.10