Parsing datashapes with "type Name = ..." in them returns None

[blaze]

Parsing datashapes with "type Name = ..." in them returns None

[mwiebe]

When I test, this, the second case now works, but the last case (ending in 3, P""") still returns None.

[sdiehl]

I'm not sure I understand the use case for defining these in one pass. The dshape() single parse is context-free so there aren't any plans to accumulate stateful definitions.

type P = {x:int32; y:int32}
3, P

Instead use:

3, { x : int32; y : int32 }

It sounds like what you want is to parse a file/raw string of datashape and return a namespace with the types. This is in the works.

[mwiebe]

I'm using this as a way to make the datashape more readable here:

https://github.com/ContinuumIO/blaze-web/blob/master/example/arrays/kiva_tiny/loans.datashape

by splitting the part which represents the actual loan data from the part which describes how the data was split up into multiple .json files in the data dump.

[sdiehl]

Yeah seems you're looking to import an external datashape file, which will be a different entry point to the parser than the dshape() which is just meant to define anonymous type declarations ala NumPy's dtype() function call.

[mwiebe]

Ok, that makes sense. I guess the right thing to do in this case is raise an exception, saying that there should just be one type declaration instead of multiple?

[sdiehl]

Yeah, seems sensible to raise an exception.

The command for loading externals will tentatively be blaze.datashape.dopen, I'll work on this for a bit.