Nested search in solr

Question

I have an Activity model and ActivityOccurrence Model where Activity has_many :activity_occurrences

Activity: This model will have all the meta data required by ActivityOccurrence

AcitvityOccurrence: attrs - occurrence(datetime), completed.

Now we have new requirement where we have to show all occurrences of activity in search results when user searches for activities in a particular range.

Previously we used to show only one record in case of repeating activities.

So as per new requirement we have decided to move search from Activity to ActivityOccurrence.

Now, I don't want to index the Meta information of Activity in each of my ActivityOccurrence as my activity has 10 fields more than ActivityOccurrence,

eg: if I have Activity with 1000 AcitivityOccurrence then I will be indexing all my activity informations in 1000 AcitivityOccurrence records.

This will take huge space as app grows if we index this way

Hence, my major concern is the amount of indexing I have to do.

So I am thinking to avoid activity indexes in ActivityOccurrence.

So is there a way to search Activity based on its filters first and then search ActivityOccurrence in the range based on the results from activities?

Note: Also we have never ending occurrences.

Any ideas?

Thanks in advance.

Answer 1

Unless you're dealing with millions of Activities/Occurrences, this may be a premature optimization - space is cheap, and SOLR is fast. Looking at this the other way around, have you considered just indexing a list of the activity occurrences that pertain to each activity (using callbacks to ensure that it gets updated)?It's hard to really optimization without more info about your data access patterns, but I'm never a fan of doing more round-trips than necessary.

That said, while I'm not sure how to write a pure SOLR query to do this, you can do it with Sunspot pretty easily:

Make sure that ActivityOccurence is searchable by Activity easily (i.e. by Activity ID).

Search Activity for the metadata that you want, and use this to extract the ID's that are relevant:

search = Activity.solr_search {<some block that does what you want>}
activity_ids = search.hits.map { |hit| hit.primary_key.to_i }

Now you can just add a with parameter to your ActivityOccurence search block:

with(:activity_id, activity_ids)

This will limit the search to the occurrences for those activities. Note that you are trading off search-time performance for index efficiency with this.