Represents a multidimensional array in a GDAL dataset. What the array says about itself is a property of it:
Details
nameandfull_name, the path from the root of the dataset.data_type_name,unit_type,nodata_value,scale,offset.data_type_class, the kind of value held:"numeric","string"or"compound".read_mdarray()reads numeric and string arrays.crs, as WKT2.dimension_count, anddimensions, a data frame with one row per dimension: itsnameandsize, thetypeanddirectionthe format gave it (both may be empty), andindexed, whether it has a coordinate variable. Dimensions are in GDAL's order, slowest varying first, which is the order a format declares them in.read_mdarray()returns itsdimthe other way round; see there for why.attributes, the format's own annotations: units, long names, conventions, valid ranges. They come back all at once because reading them one at a time is what makes inspecting a large NetCDF slow.coordinate_variables, the arrays a format names as this one's coordinates. That is not quite the same question as which dimensions are indexed: a swath carries latitude and longitude arrays that are two-dimensional and index no dimension at all.dimension_values, each dimension's coordinate variable read in full, in the array's own dimension order. For a gridded array this is the time, latitude and longitude the values are placed at, and it is the shorter road thancoordinate_variables. A dimension with no coordinate variable is NULL.block_size, how the format stores the array: the chunk shape, one value per dimension in GDAL's order and named likedimensions. A dimension the format does not chunk along is 0, as a contiguous netCDF variable is along every dimension. Reading whole chunks is what a store is fastest at, so this is the grid a chunked reader such asas_altarr()plans on.