Maybe support hdf4

[martindurant]

Working:

• find all file tags
• several informational tags
• follow LINKED lists
• find array CHUNK table references
• follow COMPressed tags
• build hierarchy

[martindurant]

This has now some interesting features

In [324]: tags, root = kerchunk.hdf.HDF4ToZarr("/Users/mdurant/data/MOD14.hdf4").translate()

In [326]: root
Out[326]: {('VG', 614)}

In [327]: tags[('VG', 614)]
Out[327]:
{'number_of_scan_lines': 2030,
 'pixels_per_scan_line': 1354,
 'number_of_active_fires': 0,
 'cmg_cells_night': 6390,
 'cmg_values': 8,
 'fire mask': {'number_of_scan_lines': 2030,
  'pixels_per_scan_line': 1354,

The two variables in here are actually "fire_mask" and "algorithm QA" in a single group, which were not shown by rasterio/xarray. The additional coordinates reported by rasterio/xarray are not in the data: "band", "spatial_ref" - I suppose rasterio wants everything to be geo-like and makes these up; there is a warning to say it is doing this:

NotGeoreferencedWarning: Dataset has no geotransform, gcps, or rpcs. The identity matrix will be returned.

[martindurant]

Latest:

In [1]: import xarray as xr
In [2]: import kerchunk.hdf
In [3]: out = kerchunk.hdf.HDF4ToZarr("/Users/mdurant/data/MOD14.hdf4").translate()
In [4]: ds = xr.open_dataset(out, engine="kerchunk")
In [5]: ds
Out[5]:
<xarray.Dataset>
Dimensions:        (CMG_night_x: 6390, CMG_night_y: 8, 0: 0,
                    algorithm QA_x: 2030, algorithm QA_y: 1354,
                    fire mask_x: 2030, fire mask_y: 1354)
Dimensions without coordinates: CMG_night_x, CMG_night_y, 0, algorithm QA_x,
                                algorithm QA_y, fire mask_x, fire mask_y
Data variables: (12/30)
    CMG_night      (CMG_night_x, CMG_night_y) float32 ...
    FP_AdjCloud    (0) float32 ...
    FP_AdjWater    (0) float32 ...
    FP_CMG_col     (0) float32 ...
    FP_CMG_row     (0) float32 ...
    FP_MAD_DT      (0) float32 ...
    ...             ...
    FP_line        (0) float32 ...
    FP_longitude   (0) float32 ...
    FP_power       (0) float32 ...
    FP_sample      (0) float32 ...
    algorithm QA   (algorithm QA_x, algorithm QA_y) float64 ...
    fire mask      (fire mask_x, fire mask_y) float32 ...
Attributes: (12/91)
    ADDITIONALATTRIBUTENAME:           identifier_product_doi_authority
    ANCILLARYINPUTPOINTER:             MOD03.A2024226.2345.061.2024227053247.hdf
    ANCILLARYINPUTTYPE:                Geolocation
    ASSOCIATEDINSTRUMENTSHORTNAME:     MODIS
    ASSOCIATEDPLATFORMSHORTNAME:       Terra
    ASSOCIATEDSENSORSHORTNAME:         MODIS
    ...                                ...
    cmg_values:                        8
    identifier_product_doi:            10.5067/MODIS/MOD14.061
    identifier_product_doi_authority:  https://doi.org
    number_of_active_fires:            0
    number_of_scan_lines:              2030
    pixels_per_scan_line:              1354