Resolution of non-literal . when README.md is in a parent directory. was: ~Environment location~

[jhheider]

Edit: the README.md is being properly located; this has become about my counter-intuitive internal mental translation of path . to the getting-started script. I should probably stop doing that.

Given how any number of tools work (npm and just spring immediately to mind), what is less surprising here:

$ tea -X git clone https://github.com/teaxyz/cli
$ cd cli
$ tea .
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   105  100   105    0     0    497      0 --:--:-- --:--:-- --:--:--   507
100 74.4M  100 74.4M    0     0  26.1M      0  0:00:02  0:00:02 --:--:-- 31.8M
tea: installed: ~/.tea/charm.sh/glow/v1.4.1

   tea really is a standalone binary  
$ cd src
$ tea .
### what do you expect here?

Current behavior is to report tea: command not found: undefined. Tools, such as those mentioned at the top, would instead start walking up the tree looking for README.md, and locating it at ... However, if we do that, should $CWD be .. or . during execution? Do we need a $SUPER_PWD within scripts, to allow for both behaviors to be available?

I'm not sure there's One True Answer to this, but I wonder what's most/least surprising to users.

[jonchang]

I imagine it would walk up the parent directory chain, stopping at filesystem mounts, looking for README.md. CWD should be the current directory. Basically, how git status works, i.e., stopping at the first .git directory found or a system mount, with the paths returned by git status (in the case of a dirty work tree) relative to the current directory.

[mxcl]

So… we already do walk up the directory tree. We stop at $HOME and / (same behavior as git).

Is something not working as expected?

[jhheider]

Hm. I seem to be getting exe/md not found for tea .. Which makes a kind of sense, since that's not the correct path to the readme. We walk for the venv, right?

[mxcl]

we walk up looking for a readme yes. I'm assuming it does not find one and then hits $HOME. Is that not what is happening?

[jhheider]

This is what I mean (using teaxyz/cli):

 ~/tea/cli $ tea .
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   105  100   105    0     0    933      0 --:--:-- --:--:-- --:--:--  1009
100 74.9M  100 74.9M    0     0  46.2M      0  0:00:01  0:00:01 --:--:-- 57.8M
tea: installed: ~/.tea/charm.sh/glow/v1.4.1
^C
 ~/tea/cli $ cd src
 ~/tea/cli/src $ tea .
tea: command not found: undefined
 ~/tea/cli/src $ tea ..
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
^C
 ~/tea/cli/src $ tea getting-started
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
^C

So, it's really only tea ., which is magical shorthand, that "doesn't work" in subdirectories. In my head, I've been translating as alias .=getting-started, but that is a different kind of confusing probably. Taken as a literal path, it makes perfect sense as implemented.

[mxcl]

K, so probs this is a bug RN, but I do think we should rethink how exe/md targets are invoked and that.