Convenient history fetch commands

[klee]

Using Sage, you are missing a feature of the Magma command line interface. In Magma, if you enter a line found in history using up arrow key and then press down arrow key, then the next line in history is fetched. This feature allows you to fetch as many successive lines in history as you like. However, Sage does not have a similar feature. The IPython command prompt uses the readline library (via pyreadline), which evidently does not support this feature. Magma has its own custom “readline-like” library, which does support this feature. (Since so many people have requested this feature, if anybody can figure out how to implement it, then such an implementation would certainly be welcome!)

This ticket addresses the above implementation.

apply trac_12503_v4.patch​

[jhpalmieri]

Is this a duplicate of #10289?

[klee]

Replying to jhpalmieri:

Is this a duplicate of #10289?

Yes. I created this because I found no way to trigger the patchbot for the patch in #10289. I hope the patchbot works for this duplicate ticket.

I will add a comment on #10289 right now.

[kcrisman]

Replying to klee:

Replying to jhpalmieri:

Is this a duplicate of #10289?

Yes. I created this because I found no way to trigger the patchbot for the patch in #10289. I hope the patchbot works for this duplicate ticket.

Just FYI, we can still give patches positive review even if the patchbot is screwing up. Maybe you should give #10289 positive review with milestone sage-duplicate? (Or maybe this one should be the duplicate, since it's a later-opened ticket.)

Also, you might want to rename the patch and change the commit message :-)

[klee]

I think I am paranoic with the patchbot! Anyway the last patch would pass the patchbot. How can I remove the first patch, and trigger the patchbot with the new patch? :-)

Please someone review this patch.

[klee]

the same

[klee]

The three patches are the same with different names. Sorry.

[klee]

changed the commit message

[klee]

apply trac_12503.patch

[klee]

Apply trac_12503.patch

[klee]

rebased on 5.0.beta12

[klee]

removed trailing spaces

[klee]

Apply trac_12503_v2.patch

[klee]

rebased on 5.0.beta12

[klee]

apply trac_12503_v3.patch

[kini]

abra cadabra!

apply trac_12503_v3.patch

voila!

[klee]

rebased on 5.1.rc0

[klee]

apply trac_12503_v3.patch

[iandrus]

I tried using this a few times before but could never bind it to Control-o so I gave up. I have an inputrc file and several other customizations that I thought might be interfering, but could never figure it out. Today I finally took the time and so I'm putting it here in case others have the same problem. Below is what happens as best I understand it. I'm on OS X 10.6.

Before readline gets a character sequence it's interpreted by the terminal emulator and can be used to do things like turn flow control on and off. These are generally bound to Control-s and Control-q and is, I think, fairly well known (at least I knew it). However, there are other actions as well which can be determined by running stty -a and reading the cchars section. In my case control-o was bound to discard. These extra bindings can be turned of by running

# Enable Ctrl-o
stty -iexten
# Enable Ctrl-s and Ctrl-q
stty -ixon

or by unbinding them individually like

stty discard undef
...

[kcrisman]

I just have to ask whether this would change current behavior in the command line. That is, would anything change for someone who didn't bind any metakeys, or didn't even really know what it was all about and was afraid of things like Emacs?

[iandrus]

Replying to kcrisman:

I just have to ask whether this would change current behavior in the command line. That is, would anything change for someone who didn't bind any metakeys, or didn't even really know what it was all about and was afraid of things like Emacs?

The current patch doesn't bind them, though I think it should. If they were bound by default (to whatever Magma uses) then they would be more discoverable, and I think the only difference (beyond perhaps a minuscule performance penalty) would be Ctl-o, which I think isn't used for anything right now. It would be good to have it confirmed that Ctl-n and Ctl-p are backwards compatible and Ctl-o is unused.

I skimmed the code and documentation and I noticed that on line 245 unbounded should be unbound. Sadly, I don't know Cython well enough to feel comfortable reviewing it, though I can say it has been working fine for me (after I got my issues sorted out). I think it would be nice to add a sentence to the documentation about stty -iexten etc.

[klee]

Replying to iandrus:

I tried using this a few times before but could never bind it to Control-o so I gave up. I have an inputrc file and several other customizations that I thought might be interfering, but could never figure it out. Today I finally took the time and so I'm putting it here in case others have the same problem. Below is what happens as best I understand it. I'm on OS X 10.6.

Did you add the relevant lines into the "ipythonrc" as indicated in the documentation? You don't modify inputrc file in your home folder.

To bind these commands with keys, insert the relevant lines into the ``$DOT_SAGE/ipython/ipythonrc``
file. Note that ``$DOT_SAGE`` is ``$HOME/.sage`` by default. For example,

::

    readline_parse_and_bind "\C-o": operate-and-get-next
    readline_parse_and_bind "\e[A": history-search-backward-and-save
    readline_parse_and_bind "\e[B": history-search-forward-and-save

binds the three commands with the control-o key, the up arrow key, and the down arrow key,
respectively.           

[klee]

Replying to kcrisman:

I just have to ask whether this would change current behavior in the command line. That is, would anything change for someone who didn't bind any metakeys, or didn't even really know what it was all about and was afraid of things like Emacs?

This patch does not chanage anything before you put some relevant lines to the "ipythonrc" file in your dot_sage folder.

[dimpase]

Replying to klee:

I can confirm that this (GAP history settings) works on OSX 10.6. Shouldn't this go into documentation?

[dimpase]

Replying to dimpase:

Replying to klee:

I can confirm that this (GAP history settings) works on OSX 10.6. Shouldn't this go into documentation?

OK, it is there already! Positive review :)

[klee]

modified to respond to the reviews

[klee]

Apply trac_12503_v4.patch

Thank you for all reviews!

[iandrus]

Replying to klee:

Replying to iandrus:

I tried using this a few times before but could never bind it to Control-o so I gave up. I have an inputrc file and several other customizations that I thought might be interfering, but could never figure it out. Today I finally took the time and so I'm putting it here in case others have the same problem. Below is what happens as best I understand it. I'm on OS X 10.6.

Did you add the relevant lines into the "ipythonrc" as indicated in the documentation? You don't modify inputrc file in your home folder.

Yes. I tried both and several other variations as well. My problem was due entirely to the fact that somewhere stty iexten (or the equivalent) was being called and so it was impossible to get the C-o binding to work in any terminal application.

[kcrisman]

I just have to ask whether this would change current behavior in the command line. That is, would anything change for someone who didn't bind any metakeys, or didn't even really know what it was all about and was afraid of things like Emacs?

This patch does not chanage anything before you put some relevant lines to the "ipythonrc" file in your dot_sage folder.

Sweet, thanks for confirming. Of course, now I might be tempted to do so...