Conversation
|
I think we can also add #176 here, looks like we currently do not support anything greater than 2D in This might also be a good time to decide if/how we want separate types for |
|
Yeah the difference between char array and string is annoying. high-dimensional char arrays are weird anyway. Currently a 3-D char array becomes a 2-D String matrix (v7.3), where the 2nd dimension is converted into the strings. I guess we could decide to only support: |
|
I don't know user preferences in Julia, so I'll refrain from commenting. Don't think its easy to get any significant opinions from end users as well is there? I'm seeing the MATLAB string type natively in datatypes like timeseries and timetable. So if we end up supporting write for those datatypes, only then we'll need a clear differentiation between chars and strings. Otherwise its a user preference issue, rather than compatibility. |
|
Personally I think the N-D char arrays are a niche MATLAB thing, I think very few people care about them, so it shouldn't matter too much. Being able to load them at all is the main priority I would say. However, we can take a moment to choose a good default now. |
I agree, this sounds good for char then. What are the possibilities for MATLAB strings? |
|
I think we can proceed after settling on a design choice that preferably does not break compatibility backwards. As I understand, we are considering the following:
This leaves out MATLAB Strings, which are right now also loaded as any of the above (depending on dimensions). It would be ideal to have some way to differentiate the two types (Char vs String), which would make it easier for saving to MAT-files. But I guess we might need to tradeoff somewhere |
I'm trying to address all the char/string array issues. Have been able to fix the v7.3 (HDF5) versions at least.
The binary version (v6, v7) still eludes me. Opening the PR to hopefully get some feedback/collaboration.
This attempts to address: #180 #143 #33