Python: Creating a multi-dimensional array out of known columns

Question

I have known columns that I want to create as an array of shape 5x269.

metadata_dtype = np.dtype([('frameNumber', np.int), ('year', np.int), ('day', np.int), ('seconds', np.int), ('sosStepIndex', np.int)])
frame_metadata_data = np.array([(frames), (years), (days), (seconds), (sos_step_index)], metadata_dtype)

frames, years, days, seconds, and sos_step_index are all 1x269 arrays. Unfortunately, when I execute the code above, I get a 5x269 array, but each item in the array contains a 1x5 list of the same repeated value.

The list of tuples is supposed to be one tuple per record. Looks like you are trying (and failing) to provide one tuple per field.. Look again at the docs and examples. — hpaulj
– hpaulj, Commented Sep 24, 2020 at 0:17
Do you have recommended docs and examples? I'm posting on Stack Overflow for the first time because I haven't been able to find an example of what I'm trying to do. — Elizabeth M
– Elizabeth M, Commented Sep 24, 2020 at 0:20
How about the numpy structured array doc page? numpy.org/doc/stable/user/basics.rec.html — hpaulj
– hpaulj, Commented Sep 24, 2020 at 0:22
Those are all examples by row, not by column. How do I convert individual columns into combined rows? — Elizabeth M
– Elizabeth M, Commented Sep 24, 2020 at 0:26

Ehsan · Accepted Answer · 2020-09-24 23:11:12Z

1

You can create an empty array and update it:

metadata_dtype = np.dtype([('frameNumber', np.int), ('year', np.int), ('day', np.int), ('seconds', np.int), ('sosStepIndex', np.int)])
frame_metadata_data = np.empty(len(frames),dtype=metadata_dtype)
frame_metadata_data['frameNumber'] = frames
frame_metadata_data['year'] = years
frame_metadata_data['day'] = days
frame_metadata_data['seconds'] = seconds
frame_metadata_data['sosStepIndex'] = sos_step_index

UPDATE: In case you want a non-structured array:

frame_metadata_data = np.stack((frames, years, days, seconds, sos_step_index)).T

In this case, you will have to have a single datatype for all arrays and will not be able to call them by names, rather you can use indices to call them. For example, frames would be frame_metadata_data[:,0]

Another approach is using Pandas:

frame_metadata_data = pd.DataFrame({'frameNumber':frames, 'year':years, 'day':days, 'seconds':seconds, 'sosStepIndex':sos_step_index})

edited Sep 24, 2020 at 23:11

answered Sep 24, 2020 at 1:00

Ehsan

12.5k2 gold badges24 silver badges36 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Elizabeth M Over a year ago

This is great in that it gets me a lot closer to what I need. Unfortunately, it seems to create a 1x256 array where each entry is the list (frameNumber, year, day, seconds, sosStepIndex). I need a 5x256 array

Ehsan Over a year ago

@ElizabethM If you do not want a structured array (which is what this post provides), you will have to have the same datatype for all your arrays (non-structured numpy arrays have single dtype). And in that case, you will not be able to name columns. I will add that approach as well as a pandas approach that might be more suitable for your case.

Collectives™ on Stack Overflow

Python: Creating a multi-dimensional array out of known columns

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related