Pull out machine-specific information to a separate config file. #618
Labels
enhancement
New feature or request
Comments
mkavulich
pushed a commit
that referenced
this issue
Jan 13, 2022
## DESCRIPTION OF CHANGES: This PR is one of two in a series that addresses refactoring platform-dependence. The topic of this PR is to refactor the logic for supporting platforms with specific real-time data feeds. Here is a list of features and modifications that were made: - Apply appropriate style for functions, mainly related to white space for nested code and comments. - Make external model file offset hours configurable for analysis- and forecast-type files because sometimes we like to start a forecast from a forecast of another model. - Refactor to reduce duplication of information. - Set up filenames and paths to be specified with templates in a consistent way for every model instead of with bash logic that is different for every model's naming convention. - Remove paths that do not exist on platforms I have access to: Jet and Hera. No changes were made to the logic for input managed by USE_USER_STAGED_EXTRN_FILES or COMINGgfs for NCO mode, although it could make sense to re-assess the NCO mode handling at a later date. I plan to go through and "review" the code to lead reviewers through this one since it bit of change. It may be helpful to view it using GitHub's ignore whitespace feature. ## TESTS CONDUCTED: Test cases using the WE2E test on Hera; see PR for full list. A test case for the same forecast configuration using known paths on Hera to exercise the new code. I checked that arrays were consistent, that the script exits in a sane manner when files are not available on disk or HPSS, and that I haven't broken anything with the way files are handled through the "user specified" mechanism necessary for the test framework. ## ISSUE (optional): This work is an incremental change in support of Issue #618 ## CONTRIBUTORS (optional): @christopherwharrop-noaa @venitahagerty @robgonzalezpita
|
Fixed via #666 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Description
I'd like to contribute a modification that pulls out as many MACHINE case statements as possible to use machine-specific configuration files that can be used during configuration and run-time so that users running on unsupported platforms do not have to modify scripts at all.
Solution
Remove as many MACHINE case statements as possible, and introduce a set of bash config files -- one for each supported machine. These will keep track of variables for machine architecture, run commands, and expected file paths.
Related to
Issue #610 -- While changes to address this issue will take into account the ability to run with a linux platform, it will still require quite a bit of script editing. For running automated testing, it will be quite difficult to specify necessary configurations to the existing tests to run on separate platforms.
The text was updated successfully, but these errors were encountered: