Mongo indexing

[natac13]

Notes

-For queries that do not use an index field, all the documents of the collection have to be scanned to fulfill the query.

• Indexes are for common queries. Therefore for what will be searched when hitting the API routes that was discussed.
• _id already is indexed
• requires 8kb of data space
  • is this a lot for our scale?
• negative performance for some write.. better to index of low write-to-read ratios
• indexes consume disk space and memory and most be tracked??
• probably looking at using compound indexes
• Need to plan the covered queries.
  • which are queries that only have criteria and projections for the indexed fields
• creating index on common queries and only using those fields in the criteria and projection mongoDB will search the index and not the collection which greatly speeds things up
• can be used on the sort() operation as well

unique indexes

• use with the sparse option as well so that docs without a value in the indexed field will be skipped over instead of getting a null insertion; which cause a duplication and therefore an error
  Created like

db.collection.createIndex({ a: 1 }, { unique: true })

Wondering if I need to do a unique index on say the email field if I specify in the Schema that that field is unique already

filtered indexed/ partial indexes

• use a filter to only include certain document in the index

db.collection.createIndex( { username: 1 }, { partialFilterExpression: { active: true } })

Only include document in the index that have the active field set to true

text indexes

• use the name option when creating so that the index name length limit is not excceed. By default mongoDB will generate a name based of all the fields in the index. WOW

Questions

• Indexing in the background.
  • So each time the app starts up and connects to mongoDB do these indexes have to be built?
  • Is it not a one time thing that gets stored in the database along with the records?
  • If so should there be unit tests that check for the creation of the indexes we desire
• Indexing stores the documents in RAM? So effectively making the queries like Redis?
  • Answer Maybe if there is a check to totalIndexSize()
  • once the value is known then there needs to be that much RAM now and for the rest of the working set??
• Clarification
• do I have the partial filter option correct?