New segment operators: "Starts with" and "Ends with"

[iMarkus]

When creating segments it is only possible to use “Is”, “Is not”, “Contains”, “Does not contain” on non-numeric values. Regarding MySQL Performance, it would make sense to implement “Starts with” and “Ends with” for segment metric definition.

[mattab]

Thanks @iMarkus for the suggestion.

[iMarkus]

I had a quick look at the code, should not be that hard to implement :)

looking forward

[tsteur]

I had a quick look at the code, should not be that hard to implement :)

We always appreciate pull requests :)

[iMarkus]

:)

[mattab]

To implement "starts with" and "ends with" we would still use the Mysql % character so I don't think it would improve performance.

@iMarkus what is the main reason you need "Starts with" and "ends with" - you write "performance" - do you have performance challenge with using "Contains"?

[iMarkus]

Hi Mattab,
I thought that using "searchstring%" and "%searchstring" would be faster than "%searchstring%".

Yes we are currently facing some performance issues on real time segments when users define their segments using "Contains" on page urls. Therefore this suggestion came up.

Regards,
Markus

[mattab]

It's true that search% may use an index, but in Piwik DB, the field log_action.name is not indexed so index won't be used (it does not exist). "Starts with" and "ends with" wouldn't speed up the query.

[iMarkus]

Alright. Thanks Matthieu for the detailed explanation. In this case it is unnecessary to implement such feature.

Best regards

[mattab]

@iMarkus Thanks for additional information too. it's always positive outcome when features don't need to be implemented :-)

[tsteur]

It looks like we will need to implement it after all as it is needed for segmentation of actionUrls in #9129 .

We'll introduce two new operators very likely: =^ for starts with and =$ for ends with

We need these operators to implement #9224