Re: [Json] Using a non-whitespace separator (Re: Working Group Last Call on draft-ietf-json-text-sequence)

[Nico Williams <[e-mail]>]

On Wed, Jun 4, 2014 at 3:36 PM, Tim Bray <[e-mail]> wrote:
> I totally don’t get why you want the LF involved.  I can’t see how it would
> make any difference, could you explain more?

Sure.

Short version:

I'm a fan of ASN.1 and PER and XDR.  But it's true that JSON is a
_textual_ encoding and so I can use $PAGER.  That's nice.  And if I'm
going to use $PAGER, then newlines between texts is... necessary.

Long version:

Let's start with this: jq(1) _today_ deals in JSON text sequences as
inputs and outputs.  You can ask jq(1) to output compact texts (like
pretty much all JSON encoders I know) or pretty-print, even colorize
texts, but jq(1) will always output a newline after every text, so
this:

% jq -C . some-file | $PAGER -R
<page through colorized pretty-printed JSON text(s) in some-file>

and the texts in 'some-file' will not run on: each will end in a newline.

Most JSON encoders I know have a "compact" output option: the whole
text with no extra whitespace.  jq(1)'s encoder does too, and so we
can:

% jq -c -f program.jq input-file > output
% wc -l output
<total number of output texts>

And sure, I could just as easily:

% jq length output
<total number of output texts>

and that's more correct even (since if 'output' came from some other
processor, there might be consecutive newlines, giving an incorrect
count with wc(1)).

But I can do all sorts of things with just text.  Like: use grep(1) as
an optimization to avoid parsing non-matching texts:

% grep -i "error.*some.error.or.other" output | jq -C -f
more-specific-filter.jq | $PAGER -R
<colorized pretty-printed texts>

Sure, it's an optimization, but don't diss optimizations.

PHB mentions $EDITOR.  I mention $PAGER.  And as I said, I'm a fan of
binary encodings.

If jq(1) spoke some binary encoding of JSON none of that would work,
but that's OK: if I had and were using such an encoding (that worked
well with sequences) I'd just not expect text-oriented tools to be any
use, or I could throw a re-encoder in the pipe so I could use
text-oriented tools.

Nico
--