Pretty Console Output --> ascii art

[sagetrac-elixyre]

My patch creates a simple ascii-art module to manipule several structure: list,dict,tuple, linear expression:

sage: R = NonCommutativeSymmetricFunctions(QQ).R()
sage: R[1]**5
R  + R   + R   + R    + R   + R    + R    + R     + R   + R    +
 *    **     *    ***     *     **      *    ****     *     **  
 *    *     **    *       *    **     ***    *        *     *   
 *    *     *     *      **    *      *               *    **   
 *    *     *            *                           **
 *

 R    + R     + R    + R     + R     + R
    *     ***      *      **       *    *****
   **    **        *    ***     ****
  **             ***

---

Apply:

• attachment: trac_14266_ascii_art_13_05_15_EliX-jbp.patch
• attachment: trac_14266-ascii_art-review-ts.patch
• attachment: trac_14266_terminal_width.patch

Depends on #8703
Depends on #14203
Depends on #14574

CC: @nthiery @VivianePons

Component: user interface

Keywords: ascii-art

Author: Jean-Baptiste Priez

Reviewer: Volker Braun, Travis Scrimshaw

Merged: sage-5.11.beta0

Issue created by migration from https://trac.sagemath.org/ticket/14266

[sagetrac-elixyre]

patch

[sagetrac-elixyre]

Attachment: pretty_console_print-EliX-jbp.patch.gz

several examples (some not in the patch)

[vbraun]

comment:1

Attachment: pcp_monkey_patch.py.gz

Whats the difference between _repr_, _pretty_repr_normal_, and _pretty_repr_term_? I think there should be only one _repr_, using ascii art unconditionally if printing the object benefits from it. Why would you want to turn it off? I don't see the need for a different display handler, we aleady have one that does handle ascii art.

[sagetrac-elixyre]

comment:2

I implement _pretty_repr_ to introduce a special ascii-art method (like in mupad) and I prefere let the method _repr_ to allow differents use. In general case, we can copy the ouput and re-use it in terminal (it is very usefull I think).
Then I put a method _pretty_repr_term_ like we have _repr_term_ for a linear expression.

I saw your handle ascii art and I try to send you a mail to merge that.... (no??).

[vbraun]

comment:3

I saw your email but #14040 has already been merged. What I'm saying is, there should be more integration into the existing framework and less reinventing the wheel. Also nobody wants an ugly representation. You can't copy/paste matrices either. To get an input that recreates the object, we already have

sage: m = matrix(ZZ, [[0, 0, 0], [0, 0, 0], [1, 0, 0], [0, 1, 0], [0, 0, 1]])
sage: m
[0 0 0]
[0 0 0]
[1 0 0]
[0 1 0]
[0 0 1]
sage: sage_input(m)
matrix(ZZ, [[0, 0, 0], [0, 0, 0], [1, 0, 0], [0, 1, 0], [0, 0, 1]])

[vbraun]

comment:4

How about a %magic to switch to the sage input representation to the displayhook, e.g.:

sage: %sage_input on
sage: m = matrix(ZZ, [[0, 0, 0], [0, 0, 0], [1, 0, 0], [0, 1, 0], [0, 0, 1]])
sage: m
matrix(ZZ, [[0, 0, 0], [0, 0, 0], [1, 0, 0], [0, 1, 0], [0, 0, 1]])

or perhaps both (if sage_input differs from _repr_)

[sagetrac-elixyre]

comment:5

I don't reinventing the wheel... Then I prefere a nice LaTeX representation too but with some objects the LaTeX compilation is very slow.
Then thanks to the method

sage: sage_input(m)

I didn't know but... after try...

sage: R = NonCommutativeSymmetricFunctions(QQ).R()
sage: a = R[1]**3; a
R  + R   + R   + R
 *    **     *    ***
 *    *     **
 *
sage: %pcp
'Pretty Console Print' is uninstalled
sage: sage_input(a)
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-10-7eccf1bd1684> in <module>()
----> 1 sage_input(a)

/home/priezjea/sage-5.8.beta4/local/lib/python2.7/site-packages/sage/misc/sage_input.pyc in sage_input(x, preparse, verify, allow_locals)
    252     if not verify:
    253         sib = SageInputBuilder(allow_locals=allow_locals, preparse=preparse)
--> 254         return sib.result(sib(x))
    255 
    256     # In verify mode, we actually compute and verify the answer with

/home/priezjea/sage-5.8.beta4/local/lib/python2.7/site-packages/sage/misc/sage_input.pyc in __call__(self, x, coerced)
    540             return SIE_literal_stringrep(self, loc_name)
    541         else:
--> 542             raise ValueError, "Can't convert %r to sage_input form"%x
    543 
    544     def preparse(self):

ValueError: Can't convert R[1, 1, 1] + R[1, 2] + R[2, 1] + R[3] to sage_input form

and

sage: R[1]**3
R[1, 1, 1] + R[1, 2] + R[2, 1] + R[3]

can be easily be copy.

So I don't say I invented something... I just purpose something to work (In my case) easily with linear combination.

[vbraun]

comment:6

So the bug is that NonCommutativeSymmetricFunctions doesn't define its own Sage input representation (define _sage_input_). The ascii art should be the default (and only) print representation.

[sagetrac-elixyre]

comment:7

I like the idea of a magic function "%sage_input on". And if every thing is ascii-art... I have some question:

1 - The _repr_ method return (and must return) a string and it is too poor to produce an efficient and usefull ascii-art output. We could image use some baseline notion or ... My opinion is an ascii-art output must be a special class like PrettyConsoleRepr or whatever which can be improve by heritage. (That is not available with string.) --> That why my opinion is we don't have to use _repr_ but anything else.

2 - How to use your code to get (easily) a list, a dict or whatever of some objects with an ascii-art output ? Your code implement only a hack for Matrix (or it was before...) no? --> If it is true my code suggests a solution (may be a bad one...)

(Sorry for my very bad english)

[VivianePons]

comment:8

I am a bit new on this discussion, but I want to add my contribution.

I think having _pretty_repr_ as a plus for combinatorial objects is a good idea and useful. It is not the same as _repr_.

First, you don't want to change the repr of every combinatorial objects to have pretty prints, this will just break everything every where. (I am thinking of trees, Dyck paths, partitions...). Most objects already had a way being pretty printed, it was just not centralized and could not be used in linear combinations.

I actually think having the two _repr_ and _pretty_repr_ is useful. I have started to work with Jean-Baptiste's patch and depending on what I want to do with the output, I use both the pretty prints and usal prints.