user search history on profile

peterjanssens

Is there a way to prune the search history for a user ?
Some users have grown that large histories and/or saved searches that they require higher max_allowed_packets config for the db, otherwise breaking user updates.

I'm not aware of any user driven pruning of those histories that probably are used to populate these

And most likely there is no system driven pruning (keep x last searches, keep x days last searches,..)
(One user has multiple quicksearches up into the thousands, that would become unusable in any UI)

Would this be a candidate for a feature request (if I'm not missing existing functionality) ? Or what would be the best way to trim here - currenly I'm looking at recreating the user and resetting ids as a quickfix.

seth

They are stored encoded in the ca_users vars field. You can write over that field with a SQL statement to remove the search history, but you'll also remove all of the user's other preferences. It will be affect archive data or anything critical, but the loss of preference settings may not be appreciated.

We should add a utility to truncate these histories, and cap their maximum length. It is on the todo list.

brucek

I will add that when backing up the database, ca_search_log is by far the largest table in the database. So much that I have started to backup all tables BUT ca_search_log, and then backing up ca_search_log separately.

I do not know how the user's search history stored in their profile relates to the ca_search_log.

I would love to be able to purge the search history (say after 1 year?, 1 month?) as I do not see too much value in the data.

Does this make sense?

seth

The search log is just a log of searches. You can safely truncate it if you don't want/need a list of searches people (or bots) have performed. You can set dont_use_search_log = 1 in app.conf to disable search logging.

peterjanssens

No, it appears to be in the ca_users variables for a user, like you can see here unserialized

https://www.unserialize.com/s/25df14d1-6eca-0629-7bb8-00000310f50d.

some of the history arrays with search results have grown quit a bit :/
(like "history";a:5386)

Maybe those lists can be truncated to x amount of past searches ?? It appears the search history dropdown(s) keep the recent ones on top, so there will some kind of timekeeping in the order of saving them (but no real timing, hence no "keep x days")
(The above dropdown is "my" search history, I cannot impersonate the impacted user to get her list - as an admin that would be a nice feature to have :p)

peterjanssens

would it be an option to have the trim here, before adding the new one ? https://github.com/collectiveaccess/providence/blob/c447a507f52af7f2f1a92a05b368376bfd67b899/app/lib/ResultContext.php#L1074

$va_history = array_slice($this->getSearchHistory(),-99);

without managing this history (per context) only grows unlimited (and stored in ca_users.vars), without real added benefits having all previous searches (and potentially outdated hits) forever ?

peterjanssens

And when we were looking at this functionality we were wondering what the added benefit of the saved searches is when a user has these (unlimited) "autosaves" ? Maybe the autosaves array can be even smaller when you have the option to (permanently) save a specific search ?
I don't think it is currently possible to manage the saved searches either, is that correct ?

peterjanssens

another candidate for curation could be "expressions_for_display" ?