call to split was taking around 60s on full
timetable file - longer than the subsequent multi-threaded section. Replacing
with a transpose of the dataframe reduces this to 3s or so. The transpose
causes the named dataframe to be coerced into a unnamed vector of int, so we
have to index into the vector by number, and convert the date coerced into
int back into a date.
call to split was taking around 60s on full timetable file - longer than the subsequent multi-threaded section. Replacing with a transpose of the dataframe reduces this to 3s or so. The transpose causes the named dataframe to be coerced into a unnamed vector of int, so we have to index into the vector by number, and convert the date coerced into int back into a date.
0001-performance(second).patch