Actually use IPython input transformer to get correct syntax

[Carreau]

This uses IPython input transformer to detect if the syntax is correct IPython.

Things that are invalid IPython do not go through : 1 ! 1 do not validate; a =! ls still does. ?,??do validate; not ???. And so on and so forth.

[jonathanslenders]

Merged: 8424287

Thank you! Really great!

[Carreau]

Thank you! Really great!

You are doing also an awesome job. Once this codebase settle a bit we could definitively discuss introducing this into IPython proper (And right now we start to be a little close to release to introduce new features). We have the requirement that IPython itself should work with zero non optional dependencies, but having multiline editting automatically if PPT is installed would be awesome.

[jonathanslenders]

Thank you! I love IPython. Having proper integration at some point would be amazing!

[Carreau]

I've put a note on the weekly hackpad to mention your project on next google hangout.
If you want to join at some later meeting we would be happy for you to join and know what your future plans are.

[jonathanslenders]

Thanks! I'd like to, it would be great to collaborate.

My current targets for this library are in priority:

• fixing existing bugs (most important.)
• making the current codebase more consistent (I'm pretty happy with the current architecture, but there are some smaller things to be improved.)
• implementing autocompletion on filenames for the system (the M-!) prompt.
• extending the Python autocompletion to complete on filenames in string literals.
• have a system to implement configurable key bindings. (I want to decouple the implementation of all the key bindings and give it names. People should be able to configure which key bindings call which commands. Also, when M-: or whatever configured is pressed, a bottom command line should also appear where commands can be entered in case some key binding is not available. I want there autocompletion as well.) This will be the hardest past, but the most fun for me.
• Implementing better line wraps (if disabled, horizontal scrolling should be possible.)
• setting up decent Sphinx documentation and have better coverage in unit tests. (Even while this is at the end, I consider it very important.)

I can't say how much time this will take. It's still a hobby project. But to give you an idea, the whole library what we have now was made in half a year.

[Carreau]

Well that seem quite reasonable. We are currently refining the completer lib in IPython, it could probably help you for some of these stuff.

[jonathanslenders]

This was not planned, but Windows support is also almost there.

I have something that works reasonable (given the limitations of the Windows console.)
But there is more testing and finishing to be done: #41

By the way, do you know what it is that causes the ctypes win32 apis to be much less fault tolorant when IPython has been loaded. I had everything working for ptpython, but when I started ptipython, I got a lot of ArgumentErrors on several calls. The good thing is that I think it's some kind of additional type checking somewhere turned on and I could fix it, but I'd very much like to know how this could be enabled.

[Carreau]

Hum, I'm not super familiar with the part where IPython use ctype for that.
Do you hate the warnings availlable somewhere ?
@minrk & @takluyver maybe know more.

[takluyver]

I can't think what we do with ctypes that should interfere with anything else. It could be something in pyreadline.

[jonathanslenders]

Well, don't worry. It's okay for me, I was just wondering.

[jonathanslenders]

@Carreau Are you coming to pycon fr this weekend? Maybe we can talk there?

[Carreau]

No, I'll be in london at MozFest, but that would have been with pleasure. Hope you will be having fun.

[jonathanslenders]

Have fun at MozFest!

[Carreau]

How was Pycon ?

[jonathanslenders]

It was nice. A lot of discussions about asyncio.