Pandas timeseries graphs for higher frequency data

[Christian Killin]

I am trying to use the plot method of series in pandas 8.0 to graph minutely data that is several years long. This can take 30 seconds to display because of the processing of the datetime index. I was thinking instead of using matplotlibs plot_date function for large series which seems to be a lot faster (even though its datetick functionality isnt as nice as pandas)The problem is that it uses an index a vector of floats representing fractional days since 010101. Is there a easy and fast way to convert a datetimeindex into the appropriate numpy array of floats? Maybe by hacking into the numpy array of datetime64s?

[Chang She]

On Sat, Jul 7, 2012 at 6:45 AM, Christian Killin <killi...@gmail.com> wrote:

I am trying to use the plot method of series in pandas 8.0 to graph minutely data that is several years long. This can take 30 seconds to display because of the processing of the datetime index. I was thinking instead of using matplotlibs plot_date function for large series which seems to be a lot faster (even though its datetick functionality isnt as nice as pandas)The problem is that it uses an index a vector of floats representing fractional days since 010101. Is there a easy and fast way to convert a datetimeindex into the appropriate numpy array of floats? Maybe by hacking into the numpy array of datetime64s?

I made a github issue here: https://github.com/pydata/pandas/issues/1579

Would you be able to post your data and a code snippet there to help us figure out the performance difference? 

Thanks,

Chang

--
Chang She

Lambda Foundry