[SERVER-18861] Queries matching null value should be fully covered by index - MongoDB Jira

XML

Word

Printable

JSON

Because of issue closing ~~SERVER-18653~~ I open a feature request for the important improvement.

Assume, we have a collectoin with documents like

{}, {a:1}, {a:[]}, {a:null}

Also we have an non sparse index on "a" field.

If we query all documents where a doesnt exist or null with {a: null}, an engine always fetches documents even having field indexed.

Need to improve query planner to allow query null valued field and to build an index which can detect collisions with array values.

depends on

SERVER-12869 Index null values and missing values differently

is duplicated by

SERVER-29326 count query with null criterion can't be covered by non-sparse index

SERVER-49845 allow indices to optionally cover missing values

SERVER-36990 Indexed filed still exists in FETCH step

is related to

SERVER-55065 Null queries should be covered by index where possible with current index format

related to

SERVER-5065 a missing field nested within an array is sorted as 'null', even if the same field exists inside another array element

SERVER-27646 covered index should be used when null inequality is in the query and index is not multikey

SERVER-12869 Index null values and missing values differently

SERVER-18653 Answering "equality to null" predicates with an index requires a FETCH stage

(4 related to)

Participants:: Alya Berciu, Andrey Hohutkin, Andrey Hvetskovich, Asya Kamsky, Backlog - Query Optimization, Charlie Swanson, Daniel Playfair Cal, David Storch, Denys Isaev, Denys Isaiev, Fuat Sungur, Kay Agahd, Ramon Fernandez Marina, Ronaldo Moura