Re: how close are we to having an "addAt" or "insertAt" feature?

[David Kastrup]

David Kastrup <address@hidden> writes:

> David Kastrup <address@hidden> writes:
>
>> Bah.  Got only 50% off, and a third of the rest is system time.  At
>> least an artificial test case went from 40 seconds to 2, but that is
>> only a 95% reduction, so still a far cry from what I fantasized.
>>
>> And the wortliste still takes more than 6 minutes on my computer:
>>
>> address@hidden:/usr/local/tmp/wortliste$ time ../git/git blame wortliste 
>> >/dev/null
>>
>> real 6m18.501s
>> user 3m58.104s
>> sys  2m17.880s
>
> Which is not even a factor of 3 as compared to previously:
>
> address@hidden:/usr/local/tmp/wortliste$ time git blame wortliste >/dev/null
>
> real  18m28.269s
> user  15m54.080s
> sys   2m27.760s

Found another worthwhile optimization making a difference (basically,
not discarding and regenerating a blob that is going to get analyzed
next anyway).  In your use case where almost all changes happen in the
same large file, this does make a bit of a difference, in particular
regarding the system time (probably mostly expended on file reads).

address@hidden:/usr/local/tmp/wortliste$ time ../git/git blame wortliste 
>/dev/null

real    4m18.314s
user    2m59.460s
sys     1m16.784s


-- 
David Kastrup