port to argparse, support reading from stdin

[ryan-williams]

fixes #1, #8

[danvk]

Nice! Can you check that this doesn't break the tests (now that I added the .spec files)?

[ryan-williams]

thanks, will do

[ryan-williams]

cool. I killed the . no-operator, which seemed not useful or no longer useful. lmk if i missed something of significance there

[danvk]

The . no-op selectors was based on jq, where it's mandatory. I believe the issue is that jss foo is ambiguous: do you want to apply the selector foo to stdin, or do you want to apply no selectors to the file foo. How does this change handle that situation?

[ryan-williams]

gotcha, i'll add it back in

[danvk]

FYI, the test failed because of some issue relating to pandoc. That dependency isn't needed for testing, so you might be able to fix it with something like:

pip install -r <(grep -v pandoc requirements.txt)

[danvk]

would be helpful to have a test for stdin input, too

[danvk]

nm about the test, I fixed it on master, you should just be able to rebase.

[ryan-williams]

ok, going back to .-semantics, what if the decision tree is:

• only 1 arg total (and no -k or -v) --> it must be a selector and input must come from stdin
• if there is a -k or -v, then there should be zero "unparsed" args (read from stdin) xor one "unparsed" arg (filename).

?

[danvk]

That sounds reasonable. What would you do with this?

jss .foo .bar .baz

is that something that's accepted now? Where does it fall in your decision tree? Is it illegal?

[danvk]

Hey Ryan -- want to get this merged in? I don't think there are any fundamental issues.

[ryan-williams]

sure, I'll take a pass on this as soon as I get my spark post out

On Wed Feb 25 2015 at 12:35:02 PM Dan Vanderkam [email protected] wrote:

Hey Ryan -- want to get this merged in? I don't think there are any fundamental issues.

— Reply to this email directly or view it on GitHub https://github.com/danvk/jss/pull/7#issuecomment-76012976.