Michael

@eric-wieser Can I just push new code and update the pull request or will this break the references in our conversation?

Just added a new version, heavily simplified, many thanks @eric-wieser!

Renamed NpyAppendArray to just AppendArray. Thanks @seberg for some inspiration on whether to keep the AppendArray class approach or extend numpy.save with an append argument (compare #11939): keeping the class...

Just send a new mail on the mailing list, see https://mail.python.org/archives/list/[email protected]/thread/A4CJ2DZCAKPMD2MYGVMDV5UI7FN4SBVI/

I have cleaned up my commits and reverted AppendArray to a state where it offers minimal functionality and tests. Should not interfere with anything else at this point and does...

Also, if we go for the `recover_trailing_content` solution mentioned above, I would suggest to add the 64 bytes of spare space in the header to every .npy array in the...

So how do we continue? Right now I have a version of AppendArray which does not affect anything else. I think follow-up pull requests should base on this one (and...

Hi Robert,yes, on append, the header size needs to be checked and the padding added accordingly. I‘ve implemented all of this in my pull request. I‘ll update it later so...

Okay, I think I got it now. I‘ll then just add some spare space to np.save so that regular .npy files can be appended to in my pull request and...

An idea came to my mind how to dynamically handle extra header padding without introducing an extra parameter: there is actually a lower and an upper bound on how long...