[netcdf-java] Reading "linearly" through a multi-dimensional array

• To: netcdf-java@xxxxxxxxxxxxxxxx
• Subject: [netcdf-java] Reading "linearly" through a multi-dimensional array
• From: Hanns Holger Rutz <rutz@xxxxxx>
• Date: Wed, 25 Jan 2017 12:20:20 +0100

hi there,

I have a scenario where I need to read the entire contents of a
multi-dimensional section'ed variable piecewise linearly.

for example, say the variable has shape [a][b][c] and I have section'ed
it as [s1][s2][s3]. say data type is double. now i need to read the
stream of elements, the total size of which is s1.size * s2.size *
s3.size in chunks, for example of chunk size x.

the API for variable doesn't allow me to specify "flat" linear indices,
so I need to calculate intermediate sub-sections, I guess. this looks
very complicated because the chunk size x might not be a simple multiple
of s3.size, s2.size*s3.size, etc.

So I can only think of two approaches: read "too" much data and then
throw away some. Read multiple times; perhaps this problem can be
formulated in maximally three read calls?

Has anyone solved this somehow? Or is there a trick in the API that I
haven't seen, to get a linear read command that preserves the offsets
and strides of sections?

thanks, .h.h.

• Follow-Ups:
  • Re: [netcdf-java] Reading "linearly" through a multi-dimensional array
    • From: Hanns Holger Rutz