Open adcroft opened 6 years ago
Are there other parts of the FRE workflow / post-processing that depend on missing_value being present? I know that had once been true.
Dr. John Krasting, Physical Scientist Biogeochemistry, Ecosystems, and Climate Group NOAA/Geophysical Fluid Dynamics Laboratory 201 Forrestal Road, Princeton, NJ 08540 P. (609) 452-5359 | F. (609) 987-5063
On Thu, May 10, 2018 at 11:26 AM, Alistair Adcroft <notifications@github.com
wrote:
A grep for missing_value returns 161 occurrences in MOM6 code. This is partly because the wrapper for register_diag_field() has missing_value as an optional argument, which we propagated upwards, but it is mainly because missing_value was briefly being used in code instead of using masks.
Apparently we have inconsistent values appearing in history files, presumably related to the specification of missing_value.
- To avoid future confusion, and inconsistent history files, we need to remove all reference to missing_value in model and framework code. Only the diag_manager should ever know anything about missing_value (technically _FillValue) which is used solely for filling space in the file where there is no data.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/MOM6/issues/769, or mute the thread https://github.com/notifications/unsubscribe-auth/AGSgc1JjSdjJKkCe2kPr5v1F50uQsUiUks5txFwOgaJpZM4T6E3U .
I do think the post-processing itself used the missing_value of _FillValue. Some of the tools used in post-processing (e.g. the nco tools) use their values. As long as they are both set in the variable's attribute, there should be no issue in the post-processing that will cause an issue.
-- Seth Underwood
S/W Development Process Lead Modeling Systems Group GFDL/NOAA/DOC 201 Forrestal Road Princeton, NJ 08540-6649
(609) 452-5847 Office (304) 376-9002 Cell (609) 987-5063 Fax Seth.Underwood@noaa.gov
On Thu, May 10, 2018 at 11:44 AM, John Krasting notifications@github.com wrote:
Are there other parts of the FRE workflow / post-processing that depend on missing_value being present? I know that had once been true.
Dr. John Krasting, Physical Scientist Biogeochemistry, Ecosystems, and Climate Group NOAA/Geophysical Fluid Dynamics Laboratory 201 Forrestal Road, Princeton, NJ 08540 https://maps.google.com/?q=201+Forrestal+Road,+Princeton,+NJ+08540&entry=gmail&source=g P. (609) 452-5359 | F. (609) 987-5063
On Thu, May 10, 2018 at 11:26 AM, Alistair Adcroft < notifications@github.com
wrote:
A grep for missing_value returns 161 occurrences in MOM6 code. This is partly because the wrapper for register_diag_field() has missing_value as an optional argument, which we propagated upwards, but it is mainly because missing_value was briefly being used in code instead of using masks.
Apparently we have inconsistent values appearing in history files, presumably related to the specification of missing_value.
- To avoid future confusion, and inconsistent history files, we need to remove all reference to missing_value in model and framework code. Only the diag_manager should ever know anything about missing_value (technically _FillValue) which is used solely for filling space in the file where there is no data.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/MOM6/issues/769, or mute the thread https://github.com/notifications/unsubscribe-auth/ AGSgc1JjSdjJKkCe2kPr5v1F50uQsUiUks5txFwOgaJpZM4T6E3U .
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/MOM6/issues/769#issuecomment-388094005, or mute the thread https://github.com/notifications/unsubscribe-auth/AAaaS8EXDr__n5GBqZPrTU8608_rbIi5ks5txGBvgaJpZM4T6E3U .
A grep for
missing_value
returns 161 occurrences in MOM6 code. This is partly because the wrapper forregister_diag_field()
hasmissing_value
as an optional argument, which we propagated upwards, but it is mainly becausemissing_value
was briefly being used in code instead of using masks.Apparently we have inconsistent values appearing in history files, presumably related to the specification of
missing_value
.missing_value
in model and framework code. Only thediag_manager
should ever know anything aboutmissing_value
(technically_FillValue
) which is used solely for filling space in the file where there is no data.