FAQ.md

Assumptions:

• Every relation is valid:
  • Relation key is present.
  • Relation target is present.

Few notes:

• $group:
  • If a $group phase is present, no $project is done.
  • Multiple $group phases probably won't work (most of the time).
• $limit:
  • Before $project (if any).
• $match:
  • If both $match and $sort are possible in the same time, $match is first.
  • A $match phase is happening as soon as any field is available.
    • We want to sort a smallest (in terms of count, not size) possible set.
• $project:
  • If $group phase is present, no $project is done.
  • Last phase (if any).
• $skip:
  • Before $project and $limit (if any).
• $sort:
  • If both $match and $sort are possible in the same time, $match is first.
  • Single $sort is happening as soon as all fields are available.
    • We want to sort a smallest (in terms of size, not count) possible set.

Ideas:

• Additional $project phase if $group present but too many fields are returned.
• Correct handling of multiple $group phases.
• Delaying $match up to some point (it might influence $group phases).
• Make $limit and $skip happen as soon as possible without changing the result.
  • This will require a note about relation arity (one-one, one-many, many-many).
• Make use of weights to choose "lighter" paths.
• Mongo 3.4 possibilities:
  • $graphLookup
• Mongo 3.6 possibilities:
  • extended $lookup usage