Hold context

[kcrisman]

Now that Pynac will support 'hold' of symbolic expressions, like

sage: tan(pi/12,hold=True)
tan(1/12*pi)

we might want a 'context' for this.
The ticket enables the following:

        sage: tan(1/12*pi)    
        -sqrt(3) + 2
        sage: with hold:
        ....:     tan(1/12*pi)
        ....:     
        tan(1/12*pi)
        
        sage: with hold:
        ....:     sin(0)
        ....:     cos(0)
        ....:     
        sin(0)
        cos(0)
        sage: hold.start()
        sage: sin(0)
        sin(0)
        sage: cos(0)
        cos(0)
        sage: hold.stop()
        sage: sin(0)
        0

Depends on #23820

CC: @burcin @eviatarbach

Component: symbolics

Author: Ralf Stephan

Branch/Commit: 3ad332c

Reviewer: Emmanuel Charpentier

Issue created by migration from https://trac.sagemath.org/ticket/10035

[jasongrout]

comment:1

Can you give a code sample of what you mean?

[kcrisman]

comment:2

Replying to @jasongrout:

Can you give a code sample of what you mean?

No, because this is really a followup to another ticket. I'll cc: Burcin in, though he probably already gets it given the component. I imagine something like

sage: context('hold') # or whatever the usual syntax is for such things - Sage doesn't have any yet
sage: a = tan(pi/12)
sage: a
tan(pi/12)

as opposed to evaluating it.

[burcin]

comment:3

I was thinking of something like:

sage: with hold:
....:     t = tan(pi/2)
....:
sage: t
tan(pi/12)
sage: tan(pi/12)
-sqrt(3) + 2

[jasongrout]

comment:4

That would be awesome!

[jpflori]

comment:5

Replying to @jasongrout:

That would be awesome!

But with such a construction it is not obvious how to simplify 't' without going through Maxima as suggested in #10034

sage: with hold: 
....:     t = tan(pi/2) 
....: 
sage: t tan(pi/12) 
sage: t.eval() #or t.unhold() or whatever 
-sqrt(3) + 2

[kcrisman]

comment:6

#11776 seems to be asking for something very similar.

[kcrisman]

comment:10

An interesting use case for this showed up at this tex.SX question.

[rwst]

comment:13

Replying to @jpflori:

Replying to @jasongrout:

That would be awesome!

But with such a construction it is not obvious how to simplify 't' without going through Maxima as suggested in #10034

sage: with hold: 
....:     t = tan(pi/2) 
....: 
sage: t tan(pi/12) 
sage: t.eval() #or t.unhold() or whatever 
-sqrt(3) + 2

You are going through Maxima because the simplify... functions do this. So improve these by replacing Maxima. It has nothing to do with the question of a global hold context.

[rwst]

comment:14

See pynac/pynac#76

[rwst]

comment:16

The construction would look like:

sage: class hold_class:
....:     def __enter__(self): print('entered')
....:     def __exit__(self, *args): print('exited')
....:     
sage: hold = hold_class()
sage: with hold:
....:     x^2
....:     
entered
x^2
exited

[rwst]

comment:17

So we can actually use a global state (e.g. a boolean hold state inside Pynac) and via the Python with ensure that it is always set locally.

[rwst]

Branch: u/rws/create_hold_context

[rwst]

Commit: 18b6c7a

[rwst]

Dependencies: pynac-0.7.11

[rwst]

Author: Ralf Stephan

[rwst]

New commits:

18b6c7a

10035: hold context

[kcrisman]

comment:20

Ordinarily I'd just say well done and leave it at that. But what to do with the "old" hold stuff? Keep, deprecate, side-by-side - should we transform old examples into this one, not, leave that to another ticket that may or may not get attention ... I assume you have thoughts on this so this is just for the purposes of hearing them, I don't have a strong opinion other than that a decision would have to be made, this patch alone wouldn't suffice. THANK YOU for this.

[rwst]

comment:21

I think the best strategy is to keep the old functionality but replace doctests in tutorials and introductory docstrings (another ticket). As there will be more such contexts a syntax for multiple settings like with context(hold, eval_fp, simplify): seems best.

Please review this whenever pynac-0.7.11 is merged.

[rwst]

comment:22

Or even with context(x > 0, simplify):

[rwst]

comment:23

Note however that the Python directive with interactively acts on its content all at once. That is,

sage: with hold:
....:     sin(0)
....:     cos(0)
....:     
sin(0)
cos(0)

Maybe we should provide start()/stop() so people can do

sage: mycontext1 = context(...)
sage: mycontext2 = context(...)
sage: with mycontext1:
........
sage: mycontext2.start()
...
sage: mycontext2.stop()

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

3ad332c

10035: hold.start/stop

[sagetrac-git]

Changed commit from 18b6c7a to 3ad332c

[rwst]

Changed dependencies from pynac-0.7.11 to #23820

[EmmanuelCharpentier]

Reviewer: Emmanuel Charpentier

[EmmanuelCharpentier]

comment:27

On top of 8.1beta9 + #24026 + #23990 + #23923 + #24088 + their dependencies running on Debian Testing on core i7 + 16 GB RAM, passes ptestlong with no error whatsoever.

==>positive_review

[rwst]

comment:28

Thanks.

[vbraun]

Changed branch from u/rws/create_hold_context to 3ad332c