Probe file should be in JSON format

[thesamovar]

No description provided.

[rossant]

@thesamovar @shabnamkadir Another option is to use YAML, a superset of JSON.

JSON example:

{
    "nchannels": 8,
    "shanks": [
        {
            "geometry": [[0, 0], [1, 1]]
        }
    ]
}

The same in YAML:

nchannels: 8
shanks:
    # Some comments
    - geometry: [[0, 0], [1, 1]]

• • More human-readable format, especially for a non-computer scientist. There are less ways of making a parse error with a missing } or ].
  • One can write comments (no comments in JSON).
• • Less widely used than JSON
  • Parsing libraries exist for most languages but are not native (need to add new external dependency). A pure Python library exists.

Here is an interesting link.

What do you think?

[thesamovar]

The benefit of JSON is its wide readability. I'd say if we were thinking about using YAML we may as well use the existing Python solution, which is just as readable. Certainly, the value stored in the HDF5 file should be in JSON I think.

[kylerbrown]

One issue with python .probe file is that, in my experience, they behave unpredictably, for example I can generate the graph like so:

probes = { 1: [(nexus32toRHD2132[a], nexus32toRHD2132[b])
               for a,b in A1x32_Poly3_50[1]]}

but the same won't work for geometry,

geometry = {nexus32toRHD2132[a]: b for a,b in A1x32_Poly3_50_geometry.items()

so I need to paste the output from a terminal.

Similarly, I generate the .params file, but I need to suppress the unicode u'prepended' to strings by casting to ascii, otherwise the program crashes, e.g.

f.write('RAW_DATA_FILES = {}\n'.format([x.encode('ascii') for x in dat_fnames]))

Therefore, I would prefer JSON or some similar format which is easy to write by hand or programatically generate.