Re: New pp (was: bug#25295: 26.0.50; Represent eieio objects using object-print in backtraces and edebug)

[Noam Postavsky <npostavs@users.sourceforge.net>]

On Sat, Mar 11, 2017 at 10:38 AM, Stefan Monnier
<monnier@iro.umontreal.ca> wrote:
>> +(defun pp-prin1 (object &optional stream)
>> +  (let ((cl-print-readably nil)
>> +        (stream (make-pp-state (or stream standard-output))))
>> +    (pp--scan :open-block stream)
>> +    (prog1 (cl-prin1 object (cons :pprint stream))
>                                ^^^^^^^^^^^^^^^^^^^^^
> Why not just pass `stream' since it's a struct and we can hence dispatch
> based on its type?

Oh, hmm, that makes sense. I took a few wrong turns while implementing
this, probably what happened is that when I decided to use (cons
:pprint) I hadn't decided on a struct yet.

>
>> +;; fallback to standard `cl-print-object'.
>> +(cl-defmethod cl-print-object (object (stream (head :pprint)))
>> +  (pp--scan (cl-prin1-to-string object) (cdr stream))
>> +  object)
>
> Hmm... but if we use such a pseudo-stream here, doesn't it break all
> calls to `princ' within other cl-print-object methods (i.e. forcing us
> the override pretty much all existing cl-print-object methods with
> a pprint-specific one)?

It doesn't break anything, AFAIK, it just means that you don't get
control over newlines vs spaces. So we would need pprint-specific
methods for "big" structures only.

Oh wait, did you possibly miss a close paren? Might be clearer this way:

(cl-defmethod cl-print-object (object (pprint-state (head :pprint)))
  (pp--scan (cl-prin1-to-string object)
            (cdr pprint-state))
  object)