Proposal for aggregation queries

proposals/0002-query-distinct.md

@@ -0,0 +1,88 @@
+# Feature name
+
+* Proposal: 0002-query-aggregate
+* Authors: [Kulshekhar Kabra](https://github.com/kulshekhar)
+* Review Manager: TBD
+* Status: **Awaiting review**
+
+## Introduction
+
+Allow users to make queries that use aggregates (distinct, sum, count, etc)
+
+## Motivation
+
+It is an often requested and useful feature which would enhance the querying capability offered by parse server
+
+One [Related issue](https://github.com/ParsePlatform/parse-server/issues/2238)
+
+## Proposed solution
+
+Introduce two additional parameters (similar to the existing `keys`, `order`, `limit`, `skip` and `include` parameters)
+
+- `group`: this will identify the fields and the aggregate functions to be applied to them, if any
+- `having`: this would be similar to the `where` clause but would be applicable to aggregated fields. 
+
+## Detailed design
+
+The new parameters will need to be whitelisted in `ClassesRouter.js`. They will also need to go through any sanitization processes that exist to prevent injection attacks.
+
+The `having` parameter without the `group` parameter would be ignored.
+
+The specification for the `group` parameter can be based on the `$group` operator from MongoDB. The operators allowed can differ based on the database adapter (The next section contains a list of operators available in Postgres and MongoDB)
+
+The specification for the `having` parameter can be based on the existing `where` parameter.
+
+The adapters will need to modify their query generation functions to incorporate these changes
+
+There is a direct relation between the new parameters introduced and Postgres's query semantics.
+
+For MongoDB the `group` parameter has a direct implementation in `$group`. To implement `having`, inspiration can be taken from [this PDF](https://rickosborne.org/download/SQL-to-MongoDB.pdf)
+
+If the `group` parameter is present in addition to the `keys` and/or `include` parameters, I'd argue that these (`keys`/`include`) should be ignored as they wouldn't really make sense. Any documentation for `group` should probably include a mention of this decision.
+
+## Aggregate operators 
+
+|Postgres|MongoDB|
+|---|---|
+||$addToSet|
+|array_agg|$push|
+|avg|$avg|
+|bit_and|
+|bit_or|
+|bool_and|
+|bool_or|
+|count|
+|every|
+||$first
+|json_agg|$push
+|jsonb_agg|$push
+|json_object_agg|
+|jsonb_object_agg|
+||$last
+|max|$max
+|min|$min
+|string_agg|
+|sum|$sum
+|xmlagg|
+|corr|
+|covar_pop|
+|covar_samp|
+|regr_avgx|
+|regr_avgy|
+|regr_count|
+|regr_intercept|
+|regr_r2|
+|regr_slope|
+|regr_sxx|
+|regr_sxy|
+|regr_syy|
+|stddev|
+|stddev_pop|$stdDevPop
+|stddev_samp|$stdDevSamp
+|variance|
+|var_pop|
+|var_samp|
+
+## Alternatives considered
+
+None really