charm searching returns odd results at times

Bug #1348290 reported by David Britton
8
This bug affects 1 person
Affects Status Importance Assigned to Milestone
charmworld
Invalid
High
Brad Crittenden
juju-gui
Fix Released
High
Jeff Pihach

Bug Description

I have attached an example from jujucharms.com that I ran just now. the unofficial, older, out-of-date, end-of-life series oneiric charm sorts higher than the official one on precise.

Is there a way that we can fix this? Note that if you try other charms you get a very interesting assortment of results back. for instance, try:

- mysql
- apache
- hadoop

Tags: landscape
Revision history for this message
Mark Shuttleworth (sabdfl) wrote :

Can I suggest this is a High priority bug? We need to steer people to the best result very effectively - they will judge us by how well the charm works, so we best steer them to the best charm.

Changed in juju-gui:
status: New → Triaged
importance: Undecided → High
affects: juju-gui → charmworld
Revision history for this message
Richard Harding (rharding) wrote :

Thanks for the bug report David. To clear, the search results are sorted appropriately. The precise/recommended charm is at the top in the top 'Recommended' results.

However, I believe this is specifically about the autocomplete results as you use the search widget on the page. The autocomplete results use the same search method, but with a different flag. We need to identify why they're prioritized correct on the page results differently than on the autocomplete ones.

Changed in charmworld:
assignee: nobody → Brad Crittenden (bac)
Revision history for this message
David Britton (dpb) wrote : Re: [Bug 1348290] Re: charm searching returns odd results at times

On Thu, Jul 24, 2014 at 09:10:57PM -0000, Richard Harding wrote:
> However, I believe this is specifically about the autocomplete results
> as you use the search widget on the page. The autocomplete results use
> the same search method, but with a different flag. We need to identify
> why they're prioritized correct on the page results differently than on
> the autocomplete ones.

Thanks Richard --

AFAICT, you are right. I just tried a barrage of charms and the
autocomplete results are the confusing point. I think we are all so
used to auto-complete results being sorted by "most important at the
top", it led to confusion at our demo.

In the case of juju-gui, more so since the auto-complete is so rich
looking. :-)

If the two algorithems could be unified easily I agree it will lead to a
more consistent experience. Thanks!

--
David Britton <email address hidden>

Brad Crittenden (bac)
Changed in charmworld:
status: Triaged → In Progress
Revision history for this message
Brad Crittenden (bac) wrote :

Both the full-text search and the autocomplete search return non-ordered results that are properly ordered by the Juju GUI. (See below.)

Since the GUI is ordering what is shown to the user for one search it should do the same for autocomplete.

http://manage.jujucharms.com/api/3/search?text=mysql&autocomplete=true
        "id": "oneiric/cf-mysql-0",
        "id": "oneiric/mysql-1",
        "id": "precise/cf-mysql-1",
        "id": "precise/mysql-46",
        "id": "trusty/mysql-3",
        "id": "~gary/demo/2/instantBigDataNoSQL",
        "id": "~francesco-chicchiricco/syncope/4/syncopePostgreSQL",
        "id": "~charmers/syncope/4/syncopePostgreSQL",
        "id": "oneiric/postgresql-0",
        "id": "precise/postgresql-74",
        "id": "trusty/postgresql-3",
        "id": "precise/postgresql-psql-7",
        "id": "~openstack-ubuntu-testing/oneiric/mysql-0",
        "id": "~openstack-ubuntu-testing/precise/mysql-5",
        "id": "~openstack-ubuntu-testing/quantal/mysql-2",
        "id": "~tribaal/trusty/mysql-1",
        "id": "~a.rosales/trusty/mysql-0",
        "id": "~bjornt/trusty/mysql-0",
        "id": "~cf-charmers/trusty/mysql-0",
        "id": "~clint-fewbar/precise/mysql-1",

http://manage.jujucharms.com/api/3/search?text=mysql
        "id": "oneiric/mysql-1",
        "id": "precise/mysql-46",
        "id": "trusty/mysql-3",
        "id": "oneiric/cf-mysql-0",
        "id": "precise/cf-mysql-1",
        "id": "oneiric/phpmyadmin-1",
        "id": "precise/phpmyadmin-4",
        "id": "precise/rails-1",
        "id": "precise/python-django-8",
        "id": "~bac/wiki/3/wiki",
        "id": "precise/lamp-5",
        "id": "~landscape/trusty/mysql-2",
        "id": "~narindergupta/trusty/mysql-0",
        "id": "~niedbalski/precise/mysql-0",
        "id": "~openstack-ubuntu-testing/oneiric/mysql-0",
        "id": "~openstack-ubuntu-testing/precise/mysql-5",
        "id": "~openstack-ubuntu-testing/quantal/mysql-2",
        "id": "~tribaal/trusty/mysql-1",
        "id": "~a.rosales/trusty/mysql-0",
        "id": "~bjornt/trusty/mysql-0",

Changed in juju-gui:
status: New → Triaged
importance: Undecided → High
Changed in charmworld:
status: In Progress → Invalid
Changed in juju-gui:
assignee: nobody → Jeff Pihach (hatch)
Brad Crittenden (bac)
Changed in juju-gui:
status: Triaged → In Progress
Revision history for this message
David Britton (dpb) wrote :
Revision history for this message
Richard Harding (rharding) wrote :

QA'ing this change on comingsoon it seems it's not corrected in the way we talked about. Goals of the change were:

- only one for each charm name would appear
- that one would be recommended, env default series, most popular (in that order)

When I try 'msyql' on comingsoon I'm still seeing multiple mysql charms that I cannot tell apart and multiple cf-mysql (and one of them shows above the recommended mysql charm)

Revision history for this message
Jeff Pihach (hatch) wrote :

The autocomplete results now show a single most relevant version of any matching charms. This makes the AC result list way more useful because the user no longer has to investigate the results they can be pretty confident that the shown 'mysql' charm is the recommended one for their environment.

Changed in juju-gui:
status: In Progress → Fix Committed
Changed in juju-gui:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.