Allow trials to associate without overriding a feature.

Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135
Committed: https://crrev.com/64e9e1144919a714e4fe3aa20f040846e2d5a9aa
Cr-Commit-Position: refs/heads/master@{#381737}

[Alexei Svitkine (slow), 2016-03-16 18:40:04]

Description was changed from

==========
Allow trials to associate without overriding a feature.

BUG=...
==========

to

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only.

More details in the code comments.

BUG=587135
==========

[Alexei Svitkine (slow), 2016-03-16 18:40:11]

Patchset #2 (id:20001) has been deleted

[Alexei Svitkine (slow), 2016-03-16 18:40:21]

Patchset #1 (id:1) has been deleted

[Alexei Svitkine (slow), 2016-03-16 18:41:10]

Description was changed from

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only.

More details in the code comments.

BUG=587135
==========

to

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135
==========

[Alexei Svitkine (slow), 2016-03-16 18:41:46]

asvitkine@chromium.org changed reviewers:
+ rkaplow@chromium.org

[rkaplow, 2016-03-17 15:08:48]

lgtm

logic seems good. I'm a bit surprised you think we should do this kind of
approach versus just enforce feature associations on all groups per study. I
understand it's a bit more work for the study designer, but I wonder if it's a
bit confusing for why it is not needed (and for example, the subtleties of it
not applying when there are multiple features).

Anyway, this seems fine.

https://codereview.chromium.org/1809633003/diff/40001/components/variations/v...
File components/variations/variations_seed_processor_unittest.cc (right):

https://codereview.chromium.org/1809633003/diff/40001/components/variations/v...
components/variations/variations_seed_processor_unittest.cc:387:
EXPECT_EQ(std::string(), processed_study.single_feature_name());
maybe note what this is testing (multiple features make s_f_n unset).

[Alexei Svitkine (slow), 2016-03-17 15:24:59]

Thanks!

I went with this approach because explicitly associating requires a proto change
(right now there's no asociated_features list - only enable/disable),
server-side config change and more burden for authors. I think this one will
cover 99.9% of cases and we can always revisit adding the other approach in the
future if we find it necessary.

https://codereview.chromium.org/1809633003/diff/40001/components/variations/v...
File components/variations/variations_seed_processor_unittest.cc (right):

https://codereview.chromium.org/1809633003/diff/40001/components/variations/v...
components/variations/variations_seed_processor_unittest.cc:387:
EXPECT_EQ(std::string(), processed_study.single_feature_name());
On 2016/03/17 15:08:48, rkaplow wrote:
> maybe note what this is testing (multiple features make s_f_n unset).

Done.

[Alexei Svitkine (slow), 2016-03-17 15:25:25]

The CQ bit was checked by asvitkine@chromium.org

[Alexei Svitkine (slow), 2016-03-17 15:25:25]

The patchset sent to the CQ was uploaded after l-g-t-m from rkaplow@chromium.org
Link to the patchset: https://codereview.chromium.org/1809633003/#ps60001
(title: "Add a comment.")

[commit-bot: I haz the power, 2016-03-17 15:25:49]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1809633003/60001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1809633003/60001

[commit-bot: I haz the power, 2016-03-17 17:32:08]

Description was changed from

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135
==========

to

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135
==========

[commit-bot: I haz the power, 2016-03-17 17:32:09]

Committed patchset #2 (id:60001)

[commit-bot: I haz the power, 2016-03-17 17:33:21]

Description was changed from

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135
==========

to

==========
Allow trials to associate without overriding a feature.

This adds a heuristic such that if a study only references a single
feature from any of its groups, its other groups will be associated
with that feature as well, for reporting only. This allows a config like
25% Enabled, 25% Disabled and 50% Default to still report the users
in the default group to UMA, so that they show up in the data.

More details in the code comments.

BUG=587135

Committed: https://crrev.com/64e9e1144919a714e4fe3aa20f040846e2d5a9aa
Cr-Commit-Position: refs/heads/master@{#381737}
==========

[commit-bot: I haz the power, 2016-03-17 17:33:22]

Patchset 2 (id:??) landed as
https://crrev.com/64e9e1144919a714e4fe3aa20f040846e2d5a9aa
Cr-Commit-Position: refs/heads/master@{#381737}