rsheftel / raccoon Goto Github PK

say I have 2 timeseries with different index

TimeSeries1

timestamp close
0 0.582778
2 0.682441
4 0.0365585
6 0.549261
8 0.0575284

TimeSeries2

timestamp open
0 0.525396
1 0.380595
2 0.399267
3 0.309334
4 0.339246
5 0.91433
6 0.30177
7 0.814401
8 0.722506
9 0.0100632

does Raccoon support the creation of Dataframe from them?
it doesn't look like it is support this but let me know if I am wrong

I was wondering if you had any plans to implement custom sorting of rows with the sort_columns function. Things like being able to pass in a custom key or comparator function, sort in reverse, etc. I don't see any way to do this currently.

.. but when going directly to https://github.com/rsheftel/raccoon/blob/master/docs/usage_dataframe.rst the code does appear

Hi there, this is a neat project that might help me with something I'm working on that doesn't lend itself well to how Pandas works. Problem is I do most of my work in Python 2.7. It didn't take long for me to make your code 2.x compatible -- mostly just replaced some of the list.copy()s with list[:]s and removed the type hints. If you're interested I can send a PR -- if not I'll just keep using what I have.

looking at get_slice():

there bisect_left from bisect is used. if blist is used as dropin it provides its own bisect_left implementation. but as far as I understand the code, this implementation is not used.

would it makes sense to query the list first if it provides a bisect function?

In set_column, I believe the following check is erroneous:

    if len(index) == (index.count(True) + index.count(False)):  # boolean list

This works in most cases, but in the rare case where the index consists only of combinations of 0 and 1, Python's quirky type system causes the code above to assume that what is being passed in is a boolean list when it's actually numeric IDs. 1==True and 0==False are helpful in many cases, but not this one.

Something like:

    if all([isinstance(i, bool) for i in index]):

Would probably do the right thing.

Currently blist is used as the backend of the array storage.
Could I ask if it is difficult enhance it with option that we can use Blosc array?
https://github.com/Blosc/bcolz

Begin the project

https://ponder.io/how-the-python-dataframe-interchange-protocol-makes-life-better/