Export to plaintext

[erik-roberts]

Allow for export to txt/csv tables, which will require unpacking of dims, in export2csv method. This is useful for using with d3js or other web frameworks. also has universal import support for other languages.

extends #40

[erik-roberts]

use csvwrite

[erik-roberts]

time series points should probably be columns to avoid unnecessary replication of the other axes.

[erik-roberts]

also, currently, unpack dims is really slow on the time dim. to speed the whole thing up, a mat array should be preallocated, and unpack dims probably shouldn't be called directly, something fast and vectorized should be called that handles all the dims at once. it would likely also use less memory in the hand offs.

[davestanley]

@erik-roberts Not sure why you need to pack / unpack. Couldn't we just do a separate file for each matrix in obj.cell?

[davestanley]

So if I understand correctly, you want to

1. Unpack all the dims in obj.data, so that time and cells (normally rows and cols of a matrix) will become 2 new axes in obj
2. Convert this to a 2D table using exportDataTable
3. Save this to a CVS?

This approach should work for MDD objects, but I would definitely advise against using it on the time dimension, since any cell array of that dimensionality is going to be uber slow.

[erik-roberts]

The full unpacking is needed for importing to other languages, especially web interfaces like d3js. Time should be columns, not rows, to avoid excess memory size.

An alternative following your idea would be to put the contents of each cell in files if they are matrices, and instead put the filename in the text output.