* Implement workaround for reading person_distinct_ids
Context under https://github.com/PostHog/product-internal/issues/114
Hopefully a temporary workaround!
* Update query
* use print
* match stickiness test
* match stickiness and trends cohort paths
* bring back distinct_id query
* revert unrelated change
* reutrn to original
* reformat
* run again
* comment out test
Co-authored-by: eric <eeoneric@gmail.com>
* add more options for funnel windowing
* tweaks
* add test
* accomodate any case and make to_dict smaller
* remove import
* change name
* store lower
* fix test
* partial implementation
* working event breakdown
* remove hardcoded fill
* add person test
* remove test unneeded
* use original breakdown implementation
* add cohort handling
* add clause lost on merge
* count display update
* do not reload list of people when loading more people
* Revert "do not reload list of people when loading more people"
This reverts commit eb9ef11690.
* fix flag
* separate initial loading state
* update title with "+", fix person/persons
Co-authored-by: Marius Andra <marius.andra@gmail.com>
* Add migration for `distinct_id` column on clickhouse
Note that code under ee/clickhouse/models/person.py does not work
anymore since it does not populate the column correctly. That is fine -
will be handled when doing the actual migration, rather than for this
experiment.
* Update plugin server to 1.1.8
Co-authored-by: posthog-bot <posthog-bot@users.noreply.github.com>
* add median time for conversion steps
* patch tests
* fix time to convert funnel query
* fix strict and unordered
* add median conversion time to base
* fix
* more tests
* fix broken funnel pagination
* account for both return results
* add limit param
* move order
* move the check
* change magic number
* adjust more tests
* Revert "Revert "Add is_deleted column to person_distinct_id (#5151)" (#5193)"
This reverts commit 401268bdba.
* A tweak for docker-compose builds
Co-authored-by: James Greenhill <fuziontech@gmail.com>
* add all steps option
* all steps working; add total and mean time to convert
* change display type checks to use enum
* kea types
* dangling console log
* Add average conversion time to time to convert results
* respond to feedabck
* responsive histogram sizes
* merged @Twixes backend changes; adjust data shape on frontend; add responsiveness to histogram
* add tooltip label
* adjust copy and tooltip
* minor tweaks
* respond to general feedback
* kea auto
* better empty state:
* error handling null time bins
* fix tests
Co-authored-by: Michael Matloka <dev@twixes.com>
Co-authored-by: Paolo D'Amico <paolodamico@users.noreply.github.com>
* Update PERSONS_ACTIVE_USER_SQL query
* Remove dead import
* Update lifecycle queries
* Update BREAKDOWN_ACTIVE_USER_INNER_SQL to use new persons query
* Update STICKINESS_SQL
* Update STICKINESS_PEOPLE_SQL
* Update STICKINESS_ACTIONS_SQL
* Update paths query
* Update events query
* Update CALCULATE_COHORT_PEOPLE_SQL
* Update retention queries
* Update TOP_PERSON_PROPS_ARRAY_OF_KEY_SQL
* Update EVENT_JOIN_PERSON_SQL
* Update GET_PERSON_ID_BY_ENTITY_COUNT_SQL
* Remove remaining references to old get latest person query
* Update GET_DISTINCT_IDS_BY_PROPERTY_SQL
* Fix code style issue
* Update table engine for person_distinct_id table
* don't select team_id
* Make person deletion work
* Use replacingmergetree over collapsing with is_deleted
Replacing an existing engine is hard, let's not do it
* Update query in test
* add migration
* set database on materialized views
* Update plugin server to 1.1.6
Co-authored-by: James Greenhill <fuziontech@gmail.com>
Co-authored-by: posthog-bot <posthog-bot@users.noreply.github.com>
* Kill dead code around updating person properties
This caused some confusion around mutations that got queued. Dead code =
bad code
* Remove unused imports
Co-authored-by: Michael Matloka <dev@twixes.com>
* Make DDLs more friendly towards running on a cluster
* Use primary CLICKHOUSE host for migrations and DDL
* loose ends on person kafka create
* posthog -> cluster typo
* add cluster to KAFKA create for plugin logs
* Feed the type monster
* clusterfy local clickhouse
* test docker-compose backed github action
* run just clickhouse and postgres from docker-compose
* move option to between up and <services>
* posthog all the things
* suggest tests run on cluster
* posthog cluster for ci
* use deploy path for docker-compose
* fix for a clickhouse bug 🐛
* complete CH bug fixes
* 5439 the github actions pg configs
* remove CLICKHOUSE_DATABASE (handled automatically)
* update DATABASE_URL for code quality checks
* Missed a few DDLs on Person
* 5439 -> 5432 to please the people
* cleanup persons and use f strings <3 f strings
* remove auto parens
* Update requirements to use our fork of infi.clickhouse_orm
* fix person.py formatting
* Include boilerplate macros for a cluster
* JOIN in funnel trends query in a more optimal way
* Reformat funnel trends assert slightly
* Refactor `ClickhouseFunnelTrends` slightly
* Add `ClickhouseFunnelTrendsPersons` code
* Allow parametrization of funnel trends persons query
* Update test_funnel_trends.py
* Clean some funnel analysis code up
* Fix `drop_off` default in `FunnelTrendsPersonsMixin`
* Refactor for `ClickhouseFunnelTrends.get_step_counts_without_aggregation_query`
* Update test_funnel_trends.py
* Add an API test for funnel trends persons
* Use `FUNNEL_PERSONS_BY_STEP_SQL`
* Tests persons some more
* Remove unused imports
* first pass mix and match everything
* cleanup
* small refactoring of get_query
* remove top level filter props
* save the file
* enable person breakdowns querying for regular + strict funnels
* move mixin testing to more appropriate location
* add unordered, testing quality of life improvements
* clean up
* add strict+unordered connectivity test for time to convert
* trim after extracting
* clean up, fix tests, trim everywhere
* Load session events asynchronously from a separate endpoint
This mirrors the behavior of postgres query
* Simplify backend & query
event_count is unused
don't select unused columns in list query
* Rename filter_by_session_recordings to filter_by_session_recordings
This is more in-line with what the function actually does
* Update types, handle start/end url properly
* start_url / end_url to session result
* Update sessions list builder tests
* Remove some `session.events` references
* Remove unneeded code
* Simplify filteredSessions
* Fix type issue
* Add test for session properties
* Test and fix start_url/end_url
* Add test for the new sessions API endpoint
* Improve types
* Update py types again
* Fix bug
