[Scipy-svn] r2560 - trunk/Lib/sandbox/timeseries/doc
scipy-svn at scipy.org
scipy-svn at scipy.org
Mon Jan 15 09:01:05 CST 2007
Date: 2007-01-15 09:01:02 -0600 (Mon, 15 Jan 2007)
New Revision: 2560
--- trunk/Lib/sandbox/timeseries/doc/todo.txt 2007-01-15 10:05:51 UTC (rev 2559)
+++ trunk/Lib/sandbox/timeseries/doc/todo.txt 2007-01-15 15:01:02 UTC (rev 2560)
@@ -4,15 +4,9 @@
but would be nice to have none-the-less :)
- - make the various Date data types actual numpy data types instead of
- an ugly hack (this is probably beyond my ability)
+ - create an actual Date data type so dates can be stored in arrays without
+ resorting to object_ dtype
- - add support for a few more frequency conversions that are missing
- in the dateOf function in tsdate.py
- - support "secondly" frequency data in a more robust way. Frequency
- conversion methods are not supported for this currently.
- add a "basis" option to the convert method. Possible values are
'business' or 'daily'. This would be used in determining the weights
of the months, quarters, etc when converting things with
@@ -21,21 +15,16 @@
This should also be an attribute of each TimeSeries object that would
provide the default behaviour if the basis option was not specified.
- - support for wider variety of standard numpy/ma functions
- add time series specific concepts/functions like moving averages, MACD, moving
standard deviation, etc...
- - add support for more frequencies: semi-annual, hourly, minutely,
- decade, weekly, and maybe even sub-frequencies like quarterly with
- quarters ending on months other than Mar/Jun/Sep/Dec, weekly ending
- on different days of the the week, etc...
- - Decide how best to handle frequency conversion between frequencies that
- don't have a nice relationship where the lower frequency strictly contains
- whole periods of the higher frequency. For example, if converting weekly to
- monthly, some weeks overlap more than one month.
+ - add support for more frequencies: semi-annual, decade, and maybe even
+ sub-frequencies like quarterly with quarters ending on months other than
+ Mar/Jun/Sep/Dec, weekly ending on different days of the the week, etc...
+ - additional unit-tests
+ - documentation
@@ -45,9 +34,10 @@
if the built in datetime module is up to the task though, particularly
where the c-api is concerned.
- - integration with pytables perhaps? Definitely a longer term thing, but eventually who knows? (I currently have my
- own module for writing this stuff to FAME databases, but pytables would be cooler)
+ - pytables submodule for reading and writing timeseries objects directly
+ to/from HDF5 databases
- - integration with matplotlib?
+ - plotting sub-module so timeseries objects can be easily plotted using
- - report generation module?
+ - reporting sub-module
More information about the Scipy-svn