Automatically update ts_created & ts_last_edit when invalidating segment reports

[9joshua]

When processing additional historical data into segments there are two steps:

1. Invalidate the Segment report
2. Edit the DB to set the segment creation date and last edit date to the beginning date of the invalidation date range

For example:
Invalidate the report...
sudo -u apache -E bash -c "php /var/www/html/console core:invalidate-report-data --sites=IDSITE_HERE --segment='SEGMENT_NAME_DEFINITION_HERE' --dates=2023-01-01,2024-03-31 --matomo-domain=domain.matomo.cloud"

Log into the DB and update the segment table...
UPDATE segment SET ts_created = '2023-01-01 00:00:00', ts_last_edit = '2023-01-01 00:00:00' WHERE name LIKE 'SEGMENT_NAME';

This process should be automatic. When the core:invalidate-report-data command is used in combination with --segment, the system should check if the invalidation start date precedes the current ts_created and ts_last_edit dates in the segment table for the requested segment name. If it does, these columns should be modified automatically to remove the need to perform step two above.

[sgiehl]

@9joshua This sound more like a workaround for processing segments beyond the configured time frame.

matomo/config/global.ini.php

Lines 315 to 323 in 826ab46

	; When archiving segments for the first time, this determines the oldest date that will be archived.
	; This option can be used to avoid archiving (for instance) the lastN years for every new segment.
	; Valid option values include: "beginning_of_time" (start date of archiving will not be changed)
	; "segment_last_edit_time" (start date of archiving will be the earliest last edit date found,
	; if none is found, the created date is used)
	; "segment_creation_time" (start date of archiving will be the creation date of the segment)
	; editLastN where N is an integer (eg "editLast10" to archive for 10 days before the segment last edit date)
	; lastN where N is an integer (eg "last10" to archive for 10 days before the segment creation date)
	process_new_segments_from = "beginning_of_time"

If this is not set to "beginning of time", segments might not be processed for the past by default. Automatically updating the ts_created of a segment would bypass this setting. So we shouldn't do that.