bug#34160: About commit bff64115a0ad081282e0f99305f41c8dd1917d67, bug#34160, json.el

[Tassilo Horn <tsdh@gnu.org>]

Lars Ingebrigtsen <larsi@gnus.org> writes:

Hi Lars,

>> It's not just moving point. replace-region-contents also keeps marks,
>> text properties and fontification intact. So we should definitely be
>> using it here.
>
> Ah, I see.  I only gave that function a cursory look-over when fixing
> this bug and I didn't quite understand what it was doing here, since I
> couldn't recall any other pretty-printer doing anything similar.

No worries!  You probably didn't see it more often because it's quite
new.  But pretty-printing is definitely a very good use-case for it.

> Sorry for the confusion here; I've now restored the
> replace-region-contents logic.  Or at least I think so; it works with
> the test cases in this bug report, at least.

Yes, it works again.  Thanks!

It can still be a bit improved in understandability and efficiency.

1. The function passed to replace-region-contents runs on the narrowed
   buffer anyway, so no need to narrow it yourself.
   
2. It would be better to create a temporary buffer, json-read repeatedly
   from the original buffer, json-encode/insert to the temp one, and
   then return the temp buffer.

The reason for point 2 is that if the function passed to
replace-region-contents returns a string, it'll put that in a temporary
buffer anyhow so that it can use replace-buffer-contents to perform the
actual replacement (replace-region-contents is just a wrapper around
that).

And we might want to cater for the situation where the region starts or
ends inside a json object by copying the buffer substring from (point)
to end to the temp buffer in case json-read fails.  I think right now,
we'd lose such half json objects and everything which follows them.

Bye,
  Tassilo