[SciPy-user] scikit.timeseries plotting bugs?

Pierre GM pgmdevlist@gmail....
Thu Dec 4 20:01:25 CST 2008


>> see also ts.extras.guess_freq
>
> I wouldn't recommend using this function as is. It is not included  
> in the docs
> right now because it isn't very robust or well tested (as opposed to  
> the usual
> reasons of laziness or lack of time :P ). I don't think it works  
> very well.

Well, we could in theory try to come up with something, following the  
example of the StringConverter (viz, you start with one freq, if this  
doesn't work go to the next one...), but don't expect it to happen any  
time soon.



>> Also I do not know whether one can create a plot from timeseries with
>> different frequencies. With reports, this wouldn't work.
>
> Plotting series of different frequencies is not supported.  
> Apparently this
> doesn't raise an error right now, but it probably should. It is easy  
> enough to
> convert series to different frequencies that I think it makes sense  
> to require
> the user to explicitly convert the series to a common frequency  
> before plotting
> rather than try to guess what the user intended.

That's actually not the way it was designed. In normal procedure, you  
create a tsplot with a time_series attached. The plot gets its  
frequency from this series, and all the series that are plotted are  
transformed to the plot frequency uisng their .asfreq method: in other  
terms, you don;t change the data, just the representation of the  
dates. It's quite convenient and I don't see why we should raise an  
exception. Issue a warning, perhaps, but let it work. <note_to_self>  
need to document that </note_to_self>



More information about the SciPy-user mailing list