Implementation/47064 support custom field in the attributesbytimestamp #13158
Conversation
dombesz
force-pushed
the
implementation/47064-support-custom_field_-in-the-attributesbytimestamp
branch
8 times, most recently
from
c1e1be4
to
0e1ec97
Compare
…lse to nil and vice versa.
dombesz
force-pushed
the
implementation/47064-support-custom_field_-in-the-attributesbytimestamp
branch
from
c77b6b6
to
a371ddf
Compare
…-the-attributesbytimestamp
There was a problem hiding this comment.
Hey Attila, I like the change. The number of places touched is really kept in check nicely.
I have two tiny remarks in the code and one issue that can also be addressed in a subsequent PR. That is actually what I would suggest, to fix the two small issues and have the bigger issue addressed separately.
The bigger issue is an N+1 query when the diff for the representer is created:
We also have https://community.openproject.org/wp/48572 where we might place the fix in. It is not strictly the same but related.
| end | ||
| end | ||
|
|
||
| def no_nil_to_empty_strings?(normalized_old_data, attribute, new_value) | ||
| old_value = normalized_old_data[attribute] | ||
| new_value != old_value && (new_value.present? || old_value.present?) | ||
| new_value != old_value |
There was a problem hiding this comment.
I thinks this method no longer serves it's purpose of checking whether value might have changed from e.g. nil to '' as the method name suggests.
new_value = nil
old_value = ''
new_value != old_value && (new_value.present? || old_value.present?)
=> false
new_value != old_value
=> true
To have the implementation be more explicit, one could change it into
([new_value, old_value] - ['', nil]).any?
Update note In an earlier version of this old_value = nil was specified which was incorrect.
There was a problem hiding this comment.
It is unsettling that no spec pics up on this. So I don't know how relevant my complaint is.
There was a problem hiding this comment.
I made the changes here 953bdf0. The expectations now fail in case any of the nil -> '' attribute transitions are included.
Some minor adjustment was required on the conditions, because [false].any? returns false and it breaks the expectation when a transition happens from nil -> false. I used ([new_value, old_value] - ['', nil]).present? instead.
| Journal | ||
| .where(id: journal_ids) | ||
| .includes({ customizable_journals: :custom_field }) | ||
| .index_by(&:id) |
There was a problem hiding this comment.
Tiny thing: The Journal needn't be involved in the loading. The same result can be achieved by
Journal::CustomizableJournal
.where(journal_id: journal_ids)
.includes(:custom_field)
.index_by(&:journal_id)
|
Thanks for the thorough review @ulferts. I addressed your remarks except the N+1 issue, that will be addressed separately. From my side this PR is good to go, once the CI is green. |
dombesz
deleted the
implementation/47064-support-custom_field_-in-the-attributesbytimestamp
branch
https://community.openproject.org/wp/47064