SigMF & MATLAB

[dxxx9]

Hello,

while I was looking for a good way to exchange IQ data, I came across SigMF.

My work consists of a Python-Matlab-Mixture, so the solution has to work for both.

When experimenting with the Matlab side I stumbled across an unfortunate Matlab "feature", as all colons (namespace separator in SigMF) will be replaced by an underscore when importing it using jsonencode().

Minimal Example

...
"core:datatype": "cf32_le", // JSON file
"antenna:azimuth_angle": 120,
...

After importing the above with jsonencode() it will become :

...
core_datatype = "cf32_le"; % in Matlab
antenna_azimuth_angle = 120;
...

This behavior is known in other contexts.

As I don't want to build an ugly wrapper for that, I wanted to ask here if anyone is working with Matlab or if there is a reasonable solution for this.

Thanks!

Daniel

PS: Only found related Issue #58.

[777arc]

Would it really be that ugly of a wrapper? I mean you'll be grabbing or inserting misc fields and it's not like core_datatype is ugly or misleading

[dxxx9]

A wrapper is definitely within the realm of possibility. But this is not the preferred solution if someone has an idea to achieve this directly. Hence the issue.

[777arc]

Oh, I didn't mean a wrapper to deal with the colon issue, I meant whatever code you are writing can be written just using underscores, like what is the actual blocker here?

[777arc]

What if you used a matlab dictionary instead of jsonencode? Pretty sure matlab keys can contain colons, they are just strings

[dxxx9]

So you suggest leaving the colon behind completely? Haven't thought about that at all. Wouldn't it manipulate the way the SigMF Metadata Format is described?

SigMF metadata fundamentally takes the form of key/value pairs: "namespace:name": value,

Even if that would only be a small adjustment, it would still mean a custom solution again, and I am trying to follow a standard.

[dxxx9]

Yes, in my attempt I use the Matlab dictionary data type to create the meta file (including a conversion to containers.Map(), as dictionaries cannot be used with jsonencode). As a minimal example that looks something like this and works fine:

metaGlobal = dictionary;
metaGlobal("core:datatype") = {"cf32_le"};
metaGlobal("core:version") = {"1.0.0"};
metaGlobal("core:sample_rate") = {1e6};

meta = struct();
meta.global = containers.Map(metaGlobal.keys, metaGlobal.values);
meta.captures = containers.Map('core:sample_start', 0);
meta.annotations = containers.Map('core:sample_start', 0);

fid = fopen("testSamples.sigmf-meta", "w");
fwrite(fid, jsonencode(meta, PrettyPrint=true));
fclose(fid);

The problem described only exists with the Matlab import, as it conflicts with the struct() data type.

[777arc]

Ah I see, is it possible to use a single dict for meta instead of a struct?

[dxxx9]

Good point! Tested some things... well, Yes and No. It's version dependent.

Before R2024b ❌

✅ Writing can be done with this modification of the above example

metaGlobal = dictionary;
metaGlobal("core:datatype") = {"cf32_le"};
metaGlobal("core:version") = {"1.0.0"};
metaGlobal("core:sample_rate") = {1e6};

meta = dictionary();
meta("global") = {containers.Map(metaGlobal.keys, metaGlobal.values)};
meta("captures") = {containers.Map("core:sample_start", 0)};
meta("annotations") = {containers.Map("core:sample_start", 0)};

fid = fopen("testSamples.sigmf-meta", "w");
fwrite(fid, jsonencode(containers.Map(meta.keys, meta.values), PrettyPrint=true));
fclose(fid);

❌ Reading is not possible as jsondecode() doesn't support Matlab's dictionary data type

With R2024b+ ✅

They introduced the writedictionary and readdictionary, so:

✅ Writing can be done with this modification of the above example

metaGlobal = dictionary;
metaGlobal("core:datatype") = {"cf32_le"};
metaGlobal("core:version") = {"1.0.0"};
metaGlobal("core:sample_rate") = {1e6};

meta = dictionary();
meta("global") = {metaGlobal};
meta("captures") = {dictionary("core:sample_start", 0)};
meta("annotations") = {dictionary("core:sample_start", 0)};

writedictionary(meta, "test_R2024b+.sigmf-meta", FileType="json");

✅ Reading can be simply done by (see colons are present)

metaImported = readdictionary("test_R2024b+.sigmf-meta", FileType="json");
metaImported{"global"}
%
% Output:
% ans =
%
%  [dictionary](matlab:helpPopup('dictionary')) (string ⟼ cell) with 3 entries:
%
%    "core:datatype"    ⟼ {["cf32_le"]}
%    "core:version"     ⟼ {["1.0.0"]}
%    "core:sample_rate" ⟼ {[1000000]}