465659 - Make the topcrasher index page not suck

Status: RESOLVED FIXED

(Socorro :: General, task)

Description

[Michael Morgan [:morgamic]]

THe topcrasher page lists all products and versions end-to-end.  This isn't very helpful.  We should lay this out better and at least group it by product and then list the most recent versions first.

Comment 1

[(not currently active) Ted Mielczarek]

A query like:
select * from branches order by product, branch desc, version desc;

Should get things in a more decent ordering. I think we should probably also add an "obsolete boolean" column to the branches table, so we can hide old versions. Then a query like:
select * from branches order by product, obsolete, branch desc, version desc;

would be even better.

Comment 2

[(not currently active) Ted Mielczarek]

Attachment: quick mockup

Something like this would be a marginal improvement. With an obsolete column we could hide old versions behind an ajax reveal, which would be super smooth.

Comment 3

[Wayne Mery (:wsmwk)]

(In reply to comment #2)
> Created an attachment (id=348960) [details]
> quick mockup
> 
> Something like this would be a marginal improvement. With an obsolete column we
> could hide old versions behind an ajax reveal, which would be super smooth.

that new ordering is nice, assuming thunderbird and seamonkey get reverse sort as well :)

Comment 4

[(not currently active) Ted Mielczarek]

Attachment: working output

I wrote a patch. Here's the output. Patch in a sec.

Comment 5

[(not currently active) Ted Mielczarek]

Attachment: unfinished patch

I implemented a version comparison function, to get real version sorting. It's sort of awful, but them's the breaks. This patch isn't complete because I changed the data structure that getProductVersions() returns, but I didn't track down and fix any other call sites. Regardless, what do you think of this implementation?

Comment 6

[(not currently active) Ted Mielczarek]

Filed bug 465820 on adding an obsolete column to branches.

Comment 7

[Michael Morgan [:morgamic]]

I'd like to do it without changing getProductVersions() since that's used in quite a few other places.

Comment 8

[Neil Lee [:neilio]]

Attachment: Attaching the original wireframe for feedback

This was wireframed up as part of the original redesign process but hasn't been implemented yet. Thoughts, suggestions, beefs, bouquets?

Comment 10

[Michael Morgan [:morgamic]]

This page still sucks.

Comment 11

[Michael Morgan [:morgamic]]

This needs to be rolled together with some nav and CSS improvements as well.  Fred - do you have time to work on this over the next week?

Comment 12

[Fred Wenzel [:wenzel]]

Yes!

Comment 13

[Fred Wenzel [:wenzel]]

Neil's original wireframe is (almost) what the front page looks like already. I personally would not let one top crashers box be on the right hand side on the frontpage, and one underneath though.

I agree that we need to change navigation a bit, currently there doesn't seem to be a fitting place for the link to the top crashes page.

Comment 14

[Fred Wenzel [:wenzel]]

Attachment: Patch, rev. 1

Here's a first patch that implements most of neilio's wireframe. The four crasher lists shown are "major", "development" and "milestone" of Firefox as well as "major" of Thunderbird.

The "other version" list is currently a list of *all* versions, which we can either leave this way or refine, if someone tells me criteria that I can filter by.

Note: When you test this, you need to go to .../topcrasher by entering it into the URL bar, there's no navigation link for it yet.

Also, ozten, if you think someone else should review this, please pass the r? on. Thanks!

Comment 15

[Fred Wenzel [:wenzel]]

Attachment: Screenshot

Here's a screenshot of the page right now.

Quick question: Can someone who knows more about crash reports than I do tell me what significance the Firefox crashes without signature have? Is my "(no signature)" correct, or would "not applicable" or something else be more appropriate?

Comment 16

[Samuel Sidler (old account; do not CC)]

"(no signature)" is more appropriate. There are varying reasons for why these signatures occur, but it's mostly because the dump isn't submitted for some weird reason.

Comment 17

[Austin King [:ozten]]

Comment on attachment 414064 [details] [diff] [review]
Patch, rev. 1

Great work! This is going to be very useful.

The patch missing a webapp-php/application/config/topcrashers.php config file. I grabbed the one from your khan instance.

Can we add one tweak? Let's use the API in webapp-php/application/libraries/versioncompare.php to sort the "Other Version" in reverse order, so newest ones will be first.

Comment 18

[Fred Wenzel [:wenzel]]

Attachment: Incremental Patch

Thanks for the comments and sorry the file was missing. This is an incremental patch, to be applied on top of the other one. I am sorting the list descendingly now.

Do we want the "featured" versions filtered from this list, by the way? If so, I could probably add that. Though maybe having a complete list at the bottom isn't so bad.

Comment 19

[Neil Lee [:neilio]]

Attachment: Tweaked mockup with 10 top crashes, select menus

Thanks wenzel, this looks super awesome.

Here's my suggestions to help clean up the "other versions" area and help expose more top crashers (tweak, ignore, critique at will, please):

- put each of the product version lists into select menus, with the top menu item being a summary of the range of versions
- make each menu (except the top-level summary item, of course) a jump menu that loads the selected versions' top crash page. (I assume we require JS so the "Go" link might not be completely necessary, but for usability's sake it might be nice to keep that there as a manual trigger.)
- centre the "View all top crashers" so it stands out a bit more, and visually reinforces that it's related to the table above
- with the saved vertical space, display the top 10 top crashers in each table.

This mockup shows all of these changes, which I believe are fairly straightforward and not too hard to implement.

Comment 20

[Fred Wenzel [:wenzel]]

I like the new mockup -- and since this page is about crashers, not versions, I agree that showing more items there and fewer at the bottom is the right thing to do.

I'll go ahead and implement that.

Comment 21

[Neil Lee [:neilio]]

Attachment: One other functionality idea - inline control of visible top crashers

Sorry, one last idea/thought - feel free to push this to another release or discard if it's not necessary.

I made two changes here:

- Moved the Product headings at the top of each top crasher table down a bit, to help visually tie the header and table together
- Added a "Show" tool that would, via ajax, reload the number of visible top crashers in the tables. The default would be 10, but you could show 20 or 50 without having to leave this page.

The thinking here is if you're tracking top crashers between the release and beta versions, being able to see more data on this page might be helpful.

Comment 22

[Fred Wenzel [:wenzel]]

Attachment: Patch, rev. 2

This is a new (not incremental) patch incorporating all of the above. It works both with and without JavaScript.

Comment 23

[Austin King [:ozten]]

Comment on attachment 414501 [details] [diff] [review]
Patch, rev. 2

Great stuff! I like the way you did the values in the select so that once you've clicked, you can use the keyboard to quickly get to the version you want. (Unlike our current query form).

Comment 24

[Fred Wenzel [:wenzel]]

Committed to revisions 1509--1514. Thanks!

Comment 25

[Austin King [:ozten]]

When you choose a product/version from the dropdown it doesn't format the url properly
Repro:
1) goto https://crash-stats.stage.mozilla.com/topcrasher/
2) Choose Firefox 3.5.3

Expected Page:
https://crash-stats.stage.mozilla.com/topcrasher/byversion/Firefox/3.5.3

Actual Page:
https://crash-stats.stage.mozilla.com/topcrasher/byversion?product=Firefox&version=3.5.3

Note:

Comment 26

[Fred Wenzel [:wenzel]]

How is that bad? The GET variables are there due to the dropdown being part of a form. The only ways I know how to circumvent is:
1) adding a redirect to the "expected page" when the "actual page" URL style was used. That'll be overhead, but it fixes the URL and works even without JS enabled.
2) rewriting the URL in JS on the fly instead of submitting the form. Doesn't work with JS disabled, obviously, and is a little dirtier.

Comment 27

[Austin King [:ozten]]

Attachment: First attempt to make noscript and REST url fix

Comment 28

[Fred Wenzel [:wenzel]]

Comment on attachment 415940 [details] [diff] [review]
First attempt to make noscript and REST url fix

Almost great.


>+	  <ol class="no_script_other_ver">
>+              <?php foreach ($versions as $version): ?>
>+	        <li><a href="<?= url::base() . 'topcrasher/byversion/' . $product . '/' . $version . '7' ?>"><?= $version ?></a></li>
>+              <?php endforeach; ?>
>+          </ol>
>+          <form method="get" action="<?php echo url::base().'topcrasher/byversion'?>" style="display: none;">

The ".'7'" needs another slash, or you end up with Firefox/3.5.1pre7 .

Also, in noscript mode, the application names are invisible. Those should stay, whether or not you have JS.

Comment 29

[Austin King [:ozten]]

Great catches. Fixed.

Sending        webapp-php/application/config/routes.php
Sending        webapp-php/application/views/topcrasher/index.php
Sending        webapp-php/js/socorro/topcrashers.js
Transmitting file data ...
Committed revision 1574.