Future `--raw-input0` and its slurping behaviour

[01mf02]

First of all: happy new year, everybody! :)

@pjungkamp has worked on an addition to jaq that implements basically --raw-input0. This is very similar to --raw-input, but splitting on '\0' instead on '\n' / "\r\n". That allows for round-tripping --raw-output0.

The interesting question is what should happen in the case of --raw-input0 --slurp. I see two options what that could do:

a) yield the original string unchanged, like --raw-input --slurp, or
b) yield an array of \0-delimited strings, like --raw-input0 -n '[inputs]'

# option a)
$ printf "abc\0def\0" | jq --raw-input0 -s
"abc\u0000def\u0000"
# option b)
$ printf "abc\0def\0" | jq --raw-input0 -s
["abc", "def"]

For me, option b) seems to be more useful in practice, but potentially also more confusing, because it yields an array of strings, whereas --raw-input --slurp yields a string.

I'm asking this question to collect feedback and to make a choice that the jq team agrees with, in case that jq wishes to implement --from-input0 as well one day.

[pjungkamp]

Let's add some cross-references to related issues:

• Allow setting raw input delimiter #965
• Feature request: support for streaming input delimited by null characters #2659

These discuss how zero-terminated string inputs could be handled, e.g., with --raw-input0.

[01mf02]

@pjungkamp, thanks for having done this research! I was not aware that there had been such concrete suggestions regarding --raw-input0 before.

Still, as far as I see, none of these suggestions have discussed the topic of --slurp.