Even when most of these are in the github issues, I leave it here so I can attack it offline (when I code best).
better error reporting, including remotes
we're weeding out imports, we could gather a list and reimport them in the remote
- from foo import bar is gonna break
imported ayrton scripts should be parsed with the ayrton parser.
becareful with if cat () | grep (); error codes must be carried too
process substitution
enable tracing
- see pdb's source
- becareful with buitins, might eclipse valid usages: bash() (exp) blocks /bin/bash
- rename bash() to expand()
- add option _exec=True for actually executing the binary.
check
bash's manpage and see what's missing.subprocess
- with ayrton (): ...
a setting for making references to unkown envvars as in bash.
trap?
executable path caching à la bash.
- what to do about relative/absolute command paths?
- git (commit) vs git.commit() vs git ('commit')
- function names are expressions too:
- / as unary op? => /path/to/excecutable and relative/path
- foo_bar vs foo__bar vs foo-bar
- -f vs (-)f vs _f
- commands in keywords should also be _out=Capture
- which is the sanest default, bash (..., single=True) or otherwise
- foo(-l, --long-option)?