Fix parentheses typo

[jasongrout]

Small typo in 0.13 What's New

[fperez]

Interesting: it wasn't really a typo, I meant it as typed. But now I realize it reads much better the other way :)

That's what I get for writing such a huge document in a maniac sprint... Oh well, we're not going for the Nobel prize in literature here...

Thanks for the fix! Merging now.

[jasongrout]

You guys are amazing. I'm really excited about this release. I've already uploaded a Sage spkg: http://trac.sagemath.org/sage_trac/ticket/12719

[fperez]

Thanks Jason, kind words much appreciated. And again, kudos for your cross-project enthusiasm, we're all the better for it!

[fperez]

BTW, regarding William's comment, that was not a mistake or omission: it was a very deliberate choice we made after lots of thought. Given that we manage numbered prompts and history, and that matplotlib produces so many plots as side effects, after much experimentation we found it worked much better to only execute with 'interactive' mode the last block.

Sage doesn't distinguish stdout from python sys.displayhook, so the opposite behavior may be more desirable, but we actually consider that a 'bad design choice' of sage, because it loses an important distinction and the cache management of output history.

So in summary: our choice is actually sensible and consistent given IPython's internal design :)

[jasongrout]

Sorry, I didn't mean to throw this simple pull request way off-topic. I'm reposting your comment to the trac ticket.

[fperez]

Sure! This is already closed, so no worries :)

[jasongrout]

I responded to your comment on the trac ticket. Reposting here just in case you don't have a trac account:

Fernando: I looked in the IPython source, and it seems that IPython is a little more subtle than mentioned above. By default, it only executes the last block as interactive (compile in 'single' mode) if it is an expression. In Sage, we (by default) execute the last block always in 'single' mode. I think that is the difference here.

Given that (and if I understood correctly), could you explain again your reasoning for choosing to not compile in 'single' mode if it isn't an expression?

[fperez]

Precisely so that if the last block has a loop, it doesn't produce multiple Out [NN] cells. We want to keep a one-to-one mapping between a cell and its output. For example it's quite common to write code like:

for i in x:
  plot(x[i], label='blah_i')

If we compile that as 'single', we'd get all the Matplotlib line collection results piling up at the bottom, one for each plot call.

We played a lot with various options, and settled on this behavior after much experimentation. After 2 years of using it, I remain convinced it's the right solution for us. Not to say that Sage can't make a different choice that works better for you guys, of course!

[jasongrout]

Thanks. That makes a lot of sense. I've posted the relevant info back to the trac ticket for Sage.

[fperez]

Awesome, thanks!

[jasongrout]

Wow, this new auto-updating of comments on pull requests is nice! It's almost like a chat.

Anyways, thanks again.

[fperez]

Yup, I love it. Been using it today extensively to effectively chat on new issues. GH rocks.

Will you make it to scipy'2012? GH folks are sponsoring and coming to give a talk. Should be fun.

[jasongrout]

No, I can't. I look forward to going another year, though.

[fperez]

Bummer. We'll be in touch regardless :)