Feature Request: Data Frames

[ghost]

Hi, In my use case, I have data sets in parquet/CSV format which I then read into a pandas dataframe for processing. Before starting to use ArrayFire Python, I would like to know if the following operations are at all supported.

1. Reading a dataframe, with its headers into an ArrayFire equivalent data structure
2. Reading a CSV, with its headers into an ArrayFire equivalent data structure
3. Writing an ArrayFire equivalent data structure into a dataframe, with its headers
4. Filtering as follows:

X_df['Rx_10G_1G'] = X_df.apply(lambda x: findGE(x['NE_OBJECT']), axis=1)  
def findGE (str_ne):    
    if str_ne.find('10GE-') !=-1:
        return 10000
    if str_ne.find('GE-') !=-1:
        return 1000
    else:
        return 1

5. Filtering as follows: X_df=X_df[X_df['Rx_Octets']> 0.0] x_neg_df=X_df[X_df['RxUtilization_pct']< 0]

6. Sorting by a time stamp based index: X_df = X_df.sort_index(by='ReportTime')

Many thanks,

[pavanky]

@QuantScientist ArrayFire does not support data frames as of now.

[ghost]

Thanks, How about np array to AF array? The documentation states:

But I get an error while trying this syntax:

[pavanky]

@QuantScientist you need to use af.to_array(array_foo)

[pavanky]

@QuantScientist Where did you find that documentation in the screenshot?

[ghost]

From "Matrix computations on the GPU with ArrayFire for Python and C/C++" Where did you take the example you quoted?

[pavanky]

@QuantScientist That's from a few years ago and outdated. Please use the documentation from here: http://arrayfire.org/arrayfire-python

Where did you take the example you quoted?

From memory :) I wrote the function.