grep: conflicting flags

[yvincent]

resolves #1256.

[petertseng]

is there a desire to match the behavior of grep? If so, it appears that this output should be iliad.txt, as I find that -l takes precedence over -n in following experiments:

$ grep -n -l hello <(echo hello)                                                              (10-08 07:58)
/proc/self/fd/11
$ grep -l -n hello <(echo hello)                                                              (10-08 07:58)
/proc/self/fd/11

[yvincent]

Yes you are right. The issue describes the opposite behavior and I didn't check. I will update the test cases accordingly.

[petertseng]

this should have a comma inside the string. Otherwise, since the flag to match entire lines and the line in the file is "Illustrious into Ades premature," the line will be included in the output.

[yvincent]

Exactly, well spotted! I'll fix that.

[petertseng]

if wanting to match the behaviour of grep, it seems this needs to be ["iliad.txt", "midsummer-night.txt", "paradise-lost.txt"], as seen in the following experiment

$ grep -n -l hello <(echo "hello\nworld") <(echo "a\nhello") <(echo "hello\nhello")
/proc/self/fd/11
/proc/self/fd/12
/proc/self/fd/13

[yvincent]

Yes, as mentioned before I followed the behavior described in the issue which was incorrect. However this should be updated to ["iliad.txt", "paradise-lost.txt"] since midsummer-night.txt doesn't contain the string "who". I'll update it as well.

[petertseng]

needs a comma, otherwise it does not match the entire line, as in https://github.com/exercism/problem-specifications/pull/1356/files#r223276145

[yvincent]

Yes, I'll fix that one too.

[petertseng]

about precedence of -l over -n, here are the grep versions I tested with that exhibited the behaviour I described:

$ grep --version
grep (BSD grep) 2.5.1-FreeBSD

and

$ grep --version
grep (GNU grep) 3.1

any grep versions with any different behaviours?

[petertseng]

I propose that this description should explicitly say which of the two flags takes precedence.

I propose this because it makes the intent of the test easier to understand for a reader of the description.

[yvincent]

It makes perfect sense. I'll update the test cases with an explicit description.

[petertseng]

this is great, I like these new descriptions!

[petertseng]

missing comma at end of this line

[petertseng]

notice that in the above multiple files tests, the lines are each prefixed with the file they appear in.

I believe the same should happen here, as demonstrated in an experiment I share below:

$ grep -x -v hello <(echo "hello\nworld") <(echo "a\nhello") <(echo "hello\nhello")
/proc/self/fd/11:world
/proc/self/fd/12:a

[petertseng]

Yup, this the new cases are matching grep's output now, just need that comma at the end of the To dogs and to all ravening fowls a prey line I pointed out now.

[yvincent]

Sorry about the missing comma, I've made the change in the wrong branch and then forgot to redo it in the correct branch. It should be all good now. Thanks!

[petertseng]

Great stuff, and obviously an improvement since it answers some reasonable questions about certain flag combinations.

Noting that this PR added the two specific flag combinations designated in the to-be-closed issue, and the discussion had not turned up any other combinations. I hope that those who see any other interesting combinations open an issue for those.

This PR has been open for almost a full week now. Anyone who has any comments on it should say so within the next 36 hours or so.

[rpottsoh]

Nice job @yvincent, thanks!

[rstockbridge]

Thank you @petertseng and @yvincent for addressing this issue!