Neuroglancer metadata datatype

[sevaa]

Looks like the tomojs-generated Neuroglancer metadata strings contain their values as string formatted integers (as opposed to numbers), like this:

{"neuroglancerPrecomputedFloor":"31403", "neuroglancerPrecomputedLimit":"34295", "neuroglancerPrecomputedMax":"33467", "neuroglancerPrecomputedMin":"32252"}

May we have compatible metadata, please? No fractions, and stored as strings. Thank you.

[blowekamp]

Please provide a reproducible example ( a small sample file, and a command line), the version of pytools and python being used. Along with the actual output, and the desired output.

Version 1.0 produces float numeric types for the value of the JSON fields.

[philipmac]

ATM the value is a float.
Am I correct in thinking @sevaa you would like that float val cast to an int, and then quoted as a string?

eg currently it's:

"metadata": {"neuroglancerPrecomputedMin": 87.22387927907181, "neuroglancerPrecomputedMax": 219.2030384667405, "neuroglancerPrecomputedFloor": 0.0, "neuroglancerPrecomputedLimit": 255.0

and you'd like:
"metadata": {"neuroglancerPrecomputedMin": "87", "neuroglancerPrecomputedMax": "219", "neuroglancerPrecomputedFloor": "0", "neuroglancerPrecomputedLimit": "255"
?

[sevaa]

Yes, precisely.

Those min/max/floor/limit parameters are used as boundaries on pixel values, and our neuroglancer pyramids all have integer pixels. I've asked Bryan, he said the devices that they had didn't emit floating point MRCs, ergo, no floating point pixels in the neuroglancer chunks, either. In tomojs I had an explicit check for that, not sure about the current workflows. Round min/floor down, max/limit up.

The fact that integers were converted to strings for storage was a quirk of tomojs that is relatively easy to replicate. If absolutely necessary, I can accommodate the datatype variance.