468258 - Can't autocomplete using unescaped UTF-8 when typing

Status: RESOLVED INCOMPLETE

(Camino Graveyard :: Location Bar & Autocomplete, defect, P3)

Description

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

STR:

1) Open http://ar.wikipedia.org/
2) Visit http://ar.wikipedia.org/wiki/مدينة_عمّان (today's feature article, on Amman)     
3) Cmd-L to get back in the location bar
4) Start typing to begin autocomplete
5) When you get done with "/wiki/", enter a م to begin completing the article name

(Substitute the Arabic wikipedia with a wikipedia with the non-Roman script of your choice.  Note also I did try with ن just in case autocomplete had the page title backwards due to busted RTL)

AR: Autocomplete selection moved to http://ar.wikipedia.org/wiki/مدينة_عمّان
ER: Autocomplete selection does not change

This is a more significant underlying issue, and in practical purposes I think it's unlikely to be hit much (but I'd like to hear from some of our non-Roman-script users/localizers, like Japanese and Russian, to be sure), but I wanted to make sure it was on file, anyway.  

FWIW, Firefox 3 appears to get this right, but I'm not sure I'm testing correctly since I'm not sure I correctly configured it to match only full url strings (and not characters in titles, or in random places in urls, etc.)

Comment 1

[philippe (part-time)]

Japanese has the same problem
STR
1. load http://ja.wikipedia.org/wiki/メインページ
2. go somewhere else
3. cmd L, start typing, once done with /wiki/, 
3a. switch to katakana input (ctrl-shift K)
3b. type 'me' (for メ), then 'return' to accept the character.

AR: no autocomplete

Minefield latest build appears to autocomplete, in my set up. And so does Safari.

(aside: can't copy paste a nice Jpn url out of Minefield location bar into  mail.app/text edit.app/textarea in Camino - works fine with Camino. yay !)

Comment 2

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

(In reply to comment #0)

> This is a more significant underlying issue, and in practical purposes I think
> it's unlikely to be hit much (but I'd like to hear from some of our
> non-Roman-script users/localizers, like Japanese and Russian, to be sure)

By which I meant to ask, how often are users going to be typing in Roman to start the autocomplete and then want to switch keyboards/input methods to finish the autocomplete using the non-Roman script part of the URL?

Comment 3

[philippe (part-time)]

(In reply to comment #2)
> (In reply to comment #0)
> 
> > This is a more significant underlying issue, and in practical purposes I think
> > it's unlikely to be hit much (but I'd like to hear from some of our
> > non-Roman-script users/localizers, like Japanese and Russian, to be sure)
> 
> By which I meant to ask, how often are users going to be typing in Roman to
> start the autocomplete and then want to switch keyboards/input methods to
> finish the autocomplete using the non-Roman script part of the URL?

Might happen: type 'wiki' (as Wikipedia is called in Japanese), let it autocomplete, delete the last part if needed and start typing.
(assuming the autocomplete after 'wiki' doesn't bring up the desired page immediately, or within the first 5 results).

I'll have to ask my partner, but I suspect she'll just go to the search box on the Wikipedia page, if the URL she's looking for doesn't pop-up immediately.

Comment 4

[philippe (part-time)]

(In reply to comment #3)

> I'll have to ask my partner, but I suspect she'll just go to the search box on
> the Wikipedia page, if the URL she's looking for doesn't pop-up immediately.

If the (Wikipedia) page doesn't pop up immediately in the dropdown: Cmd-option F (with Wikipedia search add-on) or google/yahoo.co.jp.

(The Wikipedia search box is not visible enough…)

Comment 5

[Christopher Henderson]

Attachment: Fix v1.0

Just escapes the URI before sending it to the autocomplete backend. Smokey, could you please review this for correct behaviour, then send it to smorgan for sr?

Comment 6

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

Comment on attachment 358634 [details] [diff] [review]
Fix v1.0

This is missing an implementation of escapedURI, so it doesn't build.

I think we should probably add to the comment to explain why we're escaping, too.

Comment 7

[Christopher Henderson]

Yeah, my bad. I'll hold off for now, and once the escapedURI method is in cvs (as part of Bug 464501) I'll revisit this patch.

Comment 8

[Christopher Henderson]

Attachment: Fix v1.1

Updated version to make the autocomplete behaviour the same as for escaped search strings (highlighting, canceling, arrow keys). Still depends on Bug 464501, but I'm uploading this patch now so I don't forget about it.

Comment 9

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

We should also be sure to test against the site from bug 309226, http://日本語.jp/

Comment 10

[Christopher Henderson]

Attachment: Fix v1.2

My work in 464501 has made this patch much simpler.

Comment 11

[Smokey Ardisson (offline for a while; not following bugs - do not email)]

These are P3 on the 2.1 list, but I think they're at the very bottom of hendy's to-do list.

Comment 12

[Stuart Morgan]

Pushing all the UTF-8 stuff out, since it's somewhat risky for this stage in the process.