Like you I skimmed this, having done a bit of work on RegEx's recently. I think the language in my case is using the PikeVM, by virtue of there being no error returned unlike the other engines, so I had to, within the limitations of the language and its copyrighted protected status, had to create some new RegEx functionality for myself, to make it easier to use as RegEx's can be plain and simple Voodoo.
I dont have the stats of how often the other engines would be used, but if most of the programming languages out there are using PikeVM, then I can see why Google has not only written their own OS for their servers, but also saved a few more clock cycles by getting other engines into action for specific situations where the PikeVM would be too slow and/or heavy on the clock cycles. I know all to well how a few extra characters in the search string can drastically slow up the pattern matching.
The proverb "take care of the pennies and the pounds will take care of themselves" definitely applies to RegEx and clock cycles. I suspect when looking back at some conversations from the 90's, its made some coders I know very rich when dealing with millions of records per second processing.
I dont have the stats of how often the other engines would be used, but if most of the programming languages out there are using PikeVM, then I can see why Google has not only written their own OS for their servers, but also saved a few more clock cycles by getting other engines into action for specific situations where the PikeVM would be too slow and/or heavy on the clock cycles. I know all to well how a few extra characters in the search string can drastically slow up the pattern matching.
The proverb "take care of the pennies and the pounds will take care of themselves" definitely applies to RegEx and clock cycles. I suspect when looking back at some conversations from the 90's, its made some coders I know very rich when dealing with millions of records per second processing.