alter calculation of offset

[david-bentley]

These changes allow the correct calculation of the message offset on both Windows and Linux.

There are three additional test errors introduced as part of these changes (see below - only tested on Linux so far). These are all related to grib_message being a string rather than an integer - I'm not particularly familiar with GRIB, so I'm not sure if this is a peculiarity of the tests or that one would expect to encounter a message as a string when loading a file.

======================================================================
ERROR: test_reduced_proxy_args (unit.test_GribWrapper.Test_deferred_proxy_args)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/tests/unit/test_GribWrapper.py", line 122, in test_reduced_proxy_args
    gw = GribWrapper(grib_message, self.grib_fh)
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/__init__.py", line 147, in __init__
    offset = gribapi.grib_get_message_offset(grib_message)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 1592, in grib_get_message_offset
    h = get_handle(msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 165, in get_handle
    h = ffi.cast("grib_handle*", msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/cffi/api.py", line 300, in cast
    return self._backend.cast(cdecl, source)
TypeError: an integer is required

======================================================================
ERROR: test_regular_proxy_args (unit.test_GribWrapper.Test_deferred_proxy_args)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/tests/unit/test_GribWrapper.py", line 113, in test_regular_proxy_args
    gw = GribWrapper(grib_message, self.grib_fh)
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/__init__.py", line 147, in __init__
    offset = gribapi.grib_get_message_offset(grib_message)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 1592, in grib_get_message_offset
    h = get_handle(msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 165, in get_handle
    h = ffi.cast("grib_handle*", msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/cffi/api.py", line 300, in cast
    return self._backend.cast(cdecl, source)
TypeError: an integer is required

======================================================================
ERROR: test_edition_1 (unit.test_GribWrapper.Test_edition)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/tests/unit/test_GribWrapper.py", line 74, in test_edition_1
    wrapper = GribWrapper(grib_message, grib_fh)
  File "/net/home/h02/dbentley/defence_applications/tda_source_code/github/iris-grib/iris_grib/__init__.py", line 147, in __init__
    offset = gribapi.grib_get_message_offset(grib_message)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 1592, in grib_get_message_offset
    h = get_handle(msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/gribapi/gribapi.py", line 165, in get_handle
    h = ffi.cast("grib_handle*", msgid)
  File "/data/users/dbentley/environments/iris-grib-dev/lib/python3.7/site-packages/cffi/api.py", line 300, in cast
    return self._backend.cast(cdecl, source)
TypeError: an integer is required

[david-bentley]

I've updated a couple of tests to mock the grib_get_message_offset function, so the number of test failures/errors matches those found in the main branch now

[wjbenfold]

Hi David, thanks for this. The cryptic message about a CLA (Contributor Licence Agreement) more more helpfully direct you here, which has information and a (quick) form that contributors to SciTools need to fill out: https://docs.google.com/forms/d/e/1FAIpQLSfd0tdE-DcJOXh8ej_7T93IizwJFYBFyRWYQOi2A8QRaKwykA/viewform

[david-bentley]

Now signed the CLA

[lbdreyer]

Hi @david-bentley, sometimes the cla-assistant doesn't pick up when it has been signed. I am going to close and reopen this PR to trigger it to run again

[trexfeathers]

@david-bentley are you free to rebase / merge in the latest from main?

[trexfeathers]

I'll put something together to make iris-grib NEP 29 compliant, which should at least address the iris=repo-main failure.

EDIT: working on #339

[trexfeathers]

@david-bentley if you bring in #339 I think this should pass the CI

[david-bentley]

Just for reference, here are the results of the code in #286 on Windows with these changes:

669134.5234870911 669134.5234870911
offset 0
669579.7834396362 669579.7834396362
offset 3713
669659.5537185669 669659.5537185669
offset 7426
669386.830329895 669386.830329895
offset 11139
667856.9562911987 667856.9562911987
offset 14852
666605.8734893799 666605.8734893799
offset 18565
666640.6002998352 666640.6002998352
offset 22278
666665.0799751282 666665.0799751282
offset 25991
666563.8387680054 666563.8387680054
offset 29704
666692.4885749817 666692.4885749817
offset 33417
666673.7871170044 666673.7871170044
offset 37130
666599.6193885803 666599.6193885803
offset 40843
...

Which is what we expect.

[trexfeathers]

@david-bentley this is great!

Thanks for your diligence, but even more thanks for your patience.