[with patch, positive review] plot.py improvements part 1: Remove all factories

[mhansen]

I replaced all of the factories with individual functions.

[jason]

apply instead of trac_3853.patch

[jason]

Extra things this patch does:

line([(0,0,0),(1,1,1)]) no longer produces a 3d line, but a 2d line from the first 2 coordinates. Use line3d to get a 3d line.

line(object) no longer tries object.plot() first, which is a very, very good thing, since it is natural to assume that line(object) will be something like a line, but object.plot() could be *anything*.

line no longer has a "coerce" argument; all input data is always coerced to float. I've changed this to use numpy to convert to float; currently, that's a bit faster, from some tests that wstein ran. As wstein mentioned on IRC, we should try to replace the float conversion with a snippet of Cython.

text used to make 3d text when a 3d point was passed; now it only makes 2d text. Use the text3d for 3d text.

[jason]

well, that coerce comment should really be: I changed this to use numpy in one case; the case where we need to separate the x-values and y-values still coerces to float no matter what. I guess the "line" function, a user-visible function, shouldn't have a coerce=False option; it's too unsafe.

[jason]

I take that back; I took back out the numpy stuff; we should just make a general cython function that ensures that a list is coerced to float.

[jason]

That last patch also streamlines a few things and makes a few things more consistent.

[jason]

Despite the name trac_3880 in the referee patch, it really is the referee patch for this ticket!

[jason]

referee patch updated to correct documentation and doctests, plus simplify code in xydata... function.

[jason]

plus throw some errors when trying to do line() or text() and get back a 3d object.

[jason]

there are still some doctest errors from the referee patch

[jason]

updated referee patch to fix some more things.

[jason]

Positive review for the initial part (mhansen's part). The referee patch adds enough functionality that it probably ought to be reviewed as well.

So, apply trac_3853-rebased-3.1.1-trac_3880.patch and then the referee patch to review/merge this ticket.

[jason]

I think now there is one doctest in plot.py that gives a weird error (it's a doctesting error, not a sage error, I think).

[mhansen]

Apply

trac_3853-rebased-3.1.1-trac_3880.patch trac_3880-referee.patch trac_3853-fixes.patch

in order.

[jwmerrill]

There was a thread about the proposed new behavior of line:

line([(0,0,0),(1,1,1)]) no longer produces a 3d line, but a 2d line from the first 2 coordinates. Use line3d to get a 3d line.

 http://groups.google.com/group/sage-devel/browse_thread/thread/6f4b382145c09546

All the comments were negative. That needs to be addressed, right?

I suspect I may be overstepping my bounds by changing this from positive review to needs work--if so, I'll accept an appropriate wrist slapping.

[mabshoff]

This looks good to me, it restores the old behavior. If something else is desired it should be dealt with via a new ticket since this code should go in and would bitrot quickly.

Cheers,

Michael

[mabshoff]

Merged trac_3853-rebased-3.1.1-trac_3880.patch, trac_3880-referee.patch and trac_3853-fixes.2.patch in Sage 3.1.2.alpha1