plausible/analytics
1
1
Fork 0
mirror of https://github.com/plausible/analytics.git synced 2024-12-24 01:54:34 +03:00
Code Issues Packages Projects Releases Wiki Activity

Make time range filter for sessions not break sample size estimation (#4394)

Browse Source
This commit is contained in:
Adrian Gruntkowski 2024-07-26 15:19:11 +02:00 committed by GitHub
parent c862f15f0c
commit e2d1714f90
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 14 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
lib/plausible/stats/sql/where_builder.ex
View File

@ -58,7 +58,20 @@ defmodule Plausible.Stats.SQL.WhereBuilder do
# Counts each _active_ session in time range even if they started before
dynamic(
[s],
s.site_id == ^site.id and s.timestamp >= ^first_datetime and s.start < ^last_datetime
# Currently, the sessions table in ClickHouse only has `start` column
# in its primary key. This means that filtering by `timestamp` is not
# considered when estimating number of returned rows from index
# for sample factor calculation. The redundant lower bound `start` condition
# ensures the lower bound time filter is still present as primary key
# condition and the sample factor estimation has minimal skew.
#
# Without it, the sample factor would be greatly overestimated for large sites,
# as query would be estimated to return _all_ rows matching other conditions
# before `start == last_datetime`.
s.site_id == ^site.id and
s.start >= ^NaiveDateTime.add(first_datetime, -7, :day) and
s.timestamp >= ^first_datetime and
s.start < ^last_datetime
)
end