This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

Title: rlcompleter: tab on empty prefix => insert spaces
Type: behavior Stage: resolved
Components: Library (Lib) Versions: Python 3.6, Python 3.4, Python 3.5
Status: closed Resolution: fixed
Dependencies: Superseder:
Assigned To: Nosy List: Martin Sekera, arigo, berker.peksag, dabeaz, ezio.melotti, larry, martin.panter, ned.deily, pitrou, python-dev, r.david.murray, yselivanov
Priority: deferred blocker Keywords: patch

Created on 2015-02-11 10:21 by arigo, last changed 2022-04-11 14:58 by admin. This issue is now closed.

File name Uploaded Description Edit
rlcompleter_tab2space.diff arigo, 2015-02-11 10:21 Patch review
rlcompleter-tab.patch Martin Sekera, 2015-03-14 22:58 fixes rlcompleter behavior on empty lines
issue23441.diff berker.peksag, 2015-07-26 16:51
issue23441_rev1.diff ned.deily, 2015-07-27 20:27 review
Messages (33)
msg235733 - (view) Author: Armin Rigo (arigo) * (Python committer) Date: 2015-02-11 10:21
In the interactive prompt:

>>> if 1:
... <press tab here>

Pressing tab here should just insert 4 spaces. It makes no sense to display all 500 possible "completions" of the empty word, and using tab to indent is a very common feature of any editor with a Python mode.

Patch attached.
msg235735 - (view) Author: Martin Panter (martin.panter) * (Python committer) Date: 2015-02-11 11:06
Main bug report: Issue 22086.

Can you say if this works when the cursor is not at the start of the entry? E.g. after an existing tab (Ctrl-V Tab on GNU readline) or space, or after an embedded newline (Ctrl-V Enter).

What happens if you press backspace after pressing tab? Do you have to backspace four times? Another options is to insert a literal tab code instead of spaces.
msg235750 - (view) Author: Armin Rigo (arigo) * (Python committer) Date: 2015-02-11 15:28
Ah, thanks, I missed there was already an issue.

The patch's logic is as follows: when pressing tab anywhere in the line, if the word to complete is empty (which might be for any number of reasons, like the cursor is after another space or a non-word character), then suggest as completion a string of N spaces, where N is a number from 1 to 4 so that the cursor ends up being aligned to a multiple of 4.  You have to press backspace several times to remove all these spaces.

Indeed, it could also suggest a single '\t' character, which would be simpler and revert exactly to the old behavior (including a single backspace to remove the whole tab).  I didn't do it only because it would typically indent to 8 characters instead of 4.  If people think it is anyway a better idea, I can submit a second (simpler) patch doing that.
msg235767 - (view) Author: Antoine Pitrou (pitrou) * (Python committer) Date: 2015-02-11 20:30
Thanks a lot for doing this. I think it's a good improvement. I prefer 4 spaces myself (tabs usually insert 8 spaces, which is too wide), but I'm open to other opinions.
msg236219 - (view) Author: R. David Murray (r.david.murray) * (Python committer) Date: 2015-02-19 15:33
Looks good to me.  I prefer 4 space indent as well, and since we have essentially deprecated the use of tabs for spacing in Python, I'd rather not reintroduce it here for the (minor) benefit of making backspacing easier.  Just think about the copy and paste consequences....

I'll miss vi's autoindent more than I will smart backspace, but python never had that to start with so I'm not asking for it :).

I presume someone could add an enhancement for smart backspace if they were motivated to do so, but I'd say that is a separate issue.
msg238102 - (view) Author: Martin Sekera (Martin Sekera) Date: 2015-03-14 22:58
Is it necessary to force a predefined tab width (4) onto the user here? I prefer 2-character tabs for example, and have all my terminals set up accordingly (setterm --regtabs 2). I presume many people prefer 8-column tabs, hence the default width in most software.

Additionally, the user shouldn't have to backspace 4 characters to remove one level of indentation. Not only is it annoying, it is also prone to errors if the user miscounts (and ends up with an IndentationError).

I've been using the attached patch (just 5 lines) to emulate the behavior familiar from unix shells: pressing tab activates the completion function, but indents if the line is empty (i.e. '' or all whitespace). It also indents properly if the cursor is moved to whitespace in front of a line, as one would expect.
msg238103 - (view) Author: Antoine Pitrou (pitrou) * (Python committer) Date: 2015-03-14 22:59
4 spaces is the recommended indent width in PEP 8. That's what most Python users will expect.
msg238107 - (view) Author: Ezio Melotti (ezio.melotti) * (Python committer) Date: 2015-03-14 23:30
+1 for 4 spaces.

Smart backspace would be nice too, but I agree it's a separate issue (unless it's trivial to implement).  Deeply-nested code in the interactive interpreter is uncommon, and the lack of smart backspace never bothered me.
msg238238 - (view) Author: Martin Sekera (Martin Sekera) Date: 2015-03-16 20:32
Copy that. Would the patch be acceptable if the '\t' was simply changed to 4 spaces?

To discuss further, it is my opinion that the interpreter should output a \t character when the TAB key is pressed, as this is the behavior of nearly every shell and interpreter (that doesn't autocomplete all the time) there is, including all versions of python. Introducing new behavior for the sake of compliance with PEP 8 (which, as it itself states, applies only to "the Python code comprising the standard library in the main Python distribution"), is in my opinion unnecessary and confusing (since to clear n indent levels the user has to backspace 4n times).
msg238240 - (view) Author: R. David Murray (r.david.murray) * (Python committer) Date: 2015-03-16 21:22
No program that I work with (and I use cli all the time) outputs a tab when I press the tab key.  Now, true, I think I had to configure vi so that that was the case, but nothing else that I remember.  Of course, most of those programs use readline and come with some sort of configuration such as what we are (belatedly) supply in python's command line.  So, to me, outputting an actual tab character would be *less* consistent with my general (modern) computing experience.  Especially since we are, in fact, talking about python now doing "autocomplete all the time" :)
msg238245 - (view) Author: Martin Sekera (Martin Sekera) Date: 2015-03-16 21:53
In the end it doesn't matter what characters end up on the terminal. What matters is the UX of not having to press backspace several times to unindent. That's sloppy design.

The issue of forcing a custom indent width on a user who might have their tab stops setup differently is another, albeit smaller, issue. I think the interpreter has absolutely no reason to do what's basically UI work (aligning tabbed text with tab-stop columns as in the first patch). That's the terminal's job.
msg238377 - (view) Author: R. David Murray (r.david.murray) * (Python committer) Date: 2015-03-18 05:09
The CLI is a UI.  We're using readline facilities (ie: "the terminal") to improve it.  And people cut and paste from the interactive terminal, so I think the presence of tab characters is a negative from that perspective as well.
msg238390 - (view) Author: Martin Sekera (Martin Sekera) Date: 2015-03-18 09:21
But tab characters are rendered by the terminal into spaces. During stdout processing, when the term encounters a \t (0x09), it inserts (into the term buffer that is displayed to the user) as many spaces (0x20) as needed to move the cursor to the nearest tab-stop (setterm --tabs will display them for you). Why do we need to duplicate this inside Python?

There are no copy&paste issues either, try it yourself: when you copy and paste tab-indented text from the terminal, your text will contain spaces instead of tabs (at whatever width you have your terminal tab stops configured for).
msg238392 - (view) Author: Martin Panter (martin.panter) * (Python committer) Date: 2015-03-18 09:36
I think it would really depend on the particular terminal. Sakura (a Unix terminal using the VTE library, like Gnome Terminal) copied the tab literally:

$ printf 'tab\ttab|sp   sp\n'
tab	tab|sp   sp  <== COPIED
$ hexdump -C
tab	tab|sp   sp  <== PASTED
00000000  74 61 62 09 74 61 62 7c  73 70 20 20 20 73 70 0a  ||sp   sp.|
msg242276 - (view) Author: R. David Murray (r.david.murray) * (Python committer) Date: 2015-04-30 16:25
Per the discussion in issue 5845, I'm setting this to release blocker to make sure it gets committed before the next release.  I'm also adding 3.4 as I think this is a bug.  Any objections to that?
msg246398 - (view) Author: David Beazley (dabeaz) Date: 2015-07-07 11:15
This is a problem that will never be fixed.  Sure, it was a release blocker in Python 3.4.

It wasn't fixed.

It is a release blocker in Python 3.5.

It won't be fixed.

They'll just tell you to indent using the spacebar as generations of typists have done for centuries.

It won't be fixed.

Why don't you just use ipython or bpython?

It won't be fixed.

Doesn't your IDE take care of this?

It won't be fixed.

By the way, backspace will never work right either.  No, that will never be fixed.

Did we mention that this will never be fixed?   

You can fix it!

Yes, you!  No, I mean you! Yes, yes, you can.

Simply edit the file Lib/ and comment out the line that does this:

      # enablerlcompleter()

Problem solved. All is well.

By the way.  This problem will never be fixed.  That is all.
msg246399 - (view) Author: Antoine Pitrou (pitrou) * (Python committer) Date: 2015-07-07 11:22
Well, there is a patch, and I think a couple core developers interested in this, so I wouldn't bet on this never getting fixed. Just leave them a bit more time.
msg246400 - (view) Author: David Beazley (dabeaz) Date: 2015-07-07 11:24
For what it's worth, I'm kind of tired having to hack every time I upgrade Python in order to avoid being shown 6000 choices when hitting tab on an empty line. It is crazy annoying.
msg246423 - (view) Author: Larry Hastings (larry) * (Python committer) Date: 2015-07-07 16:10
This issue was created 2015/02/11.  Python 3.4 was released on 2014/03/16.  Is there an earlier duplicate issue for this problem that was marked "release blocker" for 3.4?
msg246424 - (view) Author: R. David Murray (r.david.murray) * (Python committer) Date: 2015-07-07 16:16
No, that's what I said when I marked this one as a release blocker: the other one *hadn't been*, which is why it didn't get in to 3.4.
msg246444 - (view) Author: Martin Panter (martin.panter) * (Python committer) Date: 2015-07-08 00:36
There are currently two patches proposed, and I presume either patch would satisfy the likes of David Beazley. The first patch is more complex (extending the Completer constructor), while the second is simple. Perhaps it is okay to commit either one, and then continue discussing and refining the behaviour afterwards?

So far it looks like the only dilemma is that Martin Sekera wants to avoid the multiple backspacing and duplicating terminal behaviour, but David Murray wants to avoid the terminal copying tabs. My weak preference is with Martin’s simpler patch and relying on the Readline and the terminal to handle questions of tab stops and copied text.
msg246481 - (view) Author: David Beazley (dabeaz) Date: 2015-07-09 09:43
Frivolity aside, I really wish this issue would get more traction and a fix.

Indentation is an important part of the Python language (obviously).  A pretty standard way to indent is to hit "tab" in whatever environment you're using to edit Python code.

Yet, at the interactive prompt, tab doesn't actually indent on a blank line. Instead, it autocompletes the builtins.  Aside from it being highly annoying (as previously mentioned), it is also an embarrassment.

Newcomers to Python will very often try things out using the stock interpreter before moving on to more sophisticated environments.  The fact that tab is broken from the get-go leaves a pretty sour impression when not even the most basic tutorial examples work at the interactive console (and keep in mind that whitespace sensitivity is probably already an issue on their minds).

Experienced Python users coming from Python 2 to Python 3 are going to find that tab is busted in Python 3.  Well, of course it's busted because everything is busted in Python 3.  "Wow, this really sucks as bad as everyone says" they'll say. 

So, with that as context, I'm really hoping I don't have to watch people use a busted tab key for another entire release cycle of Python 3 as I did for Python-3.4. 

I have no particular thoughts about the specifics (tabs vs. spaces) or the amount of indentation.   It's the autocomplete on empty line that's the issue.
msg246483 - (view) Author: David Beazley (dabeaz) Date: 2015-07-09 11:11
Wanted to add:  I see this as being about the same as having a broken window pane on the front of Python 3.  Maybe there are awesome things inside, but it makes a bad first impression on anyone who dares to use the interactive console.
msg246670 - (view) Author: Antoine Pitrou (pitrou) * (Python committer) Date: 2015-07-12 20:36
> My weak preference is with Martin’s simpler patch and relying on the
> Readline and the terminal to handle questions of tab stops and copied 
> text.

Fair enough. Still, it would be nice to add a test to test_rlcompleter. Does someone want to do this?
msg247435 - (view) Author: David Beazley (dabeaz) Date: 2015-07-26 16:08
It's still broken on Python 3.5b4.
msg247438 - (view) Author: Berker Peksag (berker.peksag) * (Python committer) Date: 2015-07-26 16:51
Here is a patch with a simple test case.
msg247477 - (view) Author: Yury Selivanov (yselivanov) * (Python committer) Date: 2015-07-27 17:46
Berker, the patch looks good to me. Please commit.

Larry, is it possible we can include this in 3.5.0?
msg247481 - (view) Author: Ned Deily (ned.deily) * (Python committer) Date: 2015-07-27 20:27
Berker, the part of the patch for does not apply cleanly.  Here's an updated version of it.  Both the fix and the test seem to work as advertised on current 3.5 tip.  It would be nice to fix this finally.
msg247482 - (view) Author: Roundup Robot (python-dev) (Python triager) Date: 2015-07-27 21:08
New changeset 82ccdf2df5ac by Berker Peksag in branch '3.4':
Issue #23441: rcompleter now prints a tab character instead of displaying

New changeset d55bdd2dc45e by Berker Peksag in branch '3.5':
Issue #23441: rcompleter now prints a tab character instead of displaying

New changeset 2715734b3378 by Berker Peksag in branch 'default':
Issue #23441: rcompleter now prints a tab character instead of displaying
msg247484 - (view) Author: Berker Peksag (berker.peksag) * (Python committer) Date: 2015-07-27 21:13
Thanks for the patch Martin and thank you all for testing and reviewing patches!
msg247486 - (view) Author: Antoine Pitrou (pitrou) * (Python committer) Date: 2015-07-27 21:20
Thank you Berker for putting an end to this.
msg254869 - (view) Author: Martin Panter (martin.panter) * (Python committer) Date: 2015-11-19 01:03
The fix here doesn’t work perfectly with the Editline (libedit) version of Readline; see Issue 25660.
msg254932 - (view) Author: Yury Selivanov (yselivanov) * (Python committer) Date: 2015-11-19 20:12
Please see my patch for Issue 25660 -- instead of returning '\t' from the completion function, I use readline API to modify the input buffer directly.
Date User Action Args
2022-04-11 14:58:12adminsetgithub: 67629
2015-11-19 20:12:34yselivanovsetmessages: + msg254932
2015-11-19 01:03:47martin.pantersetmessages: + msg254869
2015-07-27 21:20:44pitrousetmessages: + msg247486
2015-07-27 21:13:46berker.peksagsetstatus: open -> closed
resolution: fixed
messages: + msg247484

stage: patch review -> resolved
2015-07-27 21:08:59python-devsetnosy: + python-dev
messages: + msg247482
2015-07-27 20:27:35ned.deilysetfiles: + issue23441_rev1.diff
nosy: + ned.deily
messages: + msg247481

2015-07-27 17:46:50yselivanovsetnosy: + yselivanov
messages: + msg247477
2015-07-26 16:51:39berker.peksagsetfiles: + issue23441.diff
versions: + Python 3.6
nosy: + berker.peksag

messages: + msg247438

stage: commit review -> patch review
2015-07-26 16:08:56dabeazsetmessages: + msg247435
2015-07-12 20:36:58pitrousetmessages: + msg246670
2015-07-09 11:11:22dabeazsetmessages: + msg246483
2015-07-09 09:43:28dabeazsetmessages: + msg246481
2015-07-08 00:36:04martin.pantersetmessages: + msg246444
2015-07-07 16:16:08r.david.murraysetmessages: + msg246424
2015-07-07 16:10:08larrysetmessages: + msg246423
2015-07-07 11:24:04dabeazsetmessages: + msg246400
2015-07-07 11:22:56pitrousetmessages: + msg246399
2015-07-07 11:15:43dabeazsetnosy: + dabeaz
messages: + msg246398
2015-07-05 01:00:58larrysetpriority: release blocker -> deferred blocker
2015-06-13 20:07:46r.david.murraylinkissue22086 superseder
2015-04-30 16:25:08r.david.murraysetpriority: normal -> release blocker

type: enhancement -> behavior
versions: + Python 3.4
nosy: + larry

messages: + msg242276
stage: patch review -> commit review
2015-03-18 09:36:52martin.pantersetmessages: + msg238392
2015-03-18 09:21:56Martin Sekerasetmessages: + msg238390
2015-03-18 05:09:47r.david.murraysetmessages: + msg238377
2015-03-16 21:53:50Martin Sekerasetmessages: + msg238245
2015-03-16 21:22:56r.david.murraysetmessages: + msg238240
2015-03-16 20:32:55Martin Sekerasetmessages: + msg238238
2015-03-14 23:30:47ezio.melottisetnosy: + ezio.melotti
messages: + msg238107
2015-03-14 22:59:13pitrousetmessages: + msg238103
2015-03-14 22:58:23Martin Sekerasetfiles: + rlcompleter-tab.patch
nosy: + Martin Sekera
messages: + msg238102

2015-02-19 15:33:55r.david.murraysetmessages: + msg236219
2015-02-11 20:30:02pitrousetversions: - Python 3.6
nosy: + r.david.murray, pitrou

messages: + msg235767

stage: patch review
2015-02-11 15:28:16arigosetmessages: + msg235750
2015-02-11 11:06:06martin.pantersetnosy: + martin.panter
messages: + msg235735
2015-02-11 10:21:44arigocreate