Check if experiment_id attribute is present in current dataset when being processed by PrePARE

[mauzey1]

Fixes #643

[durack1]

@mauzey1 thanks for catching this - @matthew-mizielinski are there any other obvious tweaks you suggest?

[piotr-florek-mohc]

Looks good to me.

[piotr-florek-mohc]

Just a quick note, this fix doesn't appear to resolve the whole issue raised in #643 entirely (at least it didn't do it for me), as there are other problems with the netCDF file, which cmor to fail because of weird characters appearing in the log file. The quick fix, as commented by @matthew.mizielinski, is to provide encoding parameter to the PrePARE.py script.

diff --git a/LibCV/PrePARE/PrePARE.py b/LibCV/PrePARE/PrePARE.py
index 6fcd570..97f1c9e 100755
--- a/LibCV/PrePARE/PrePARE.py
+++ b/LibCV/PrePARE/PrePARE.py
@@ -931,7 +931,7 @@ def main():
         remove_ansi = re.compile(r'\x1b\[[0-?]*[ -/]*[@-~]')
         for logfile in set(logfiles):
             if not os.stat(logfile).st_size == 0:
-                with open(logfile, 'r') as f:
+                with open(logfile, 'r', encoding='utf8', errors='ignore') as f:
                     log_text = f.read()
                     if args.no_text_color:
                         log_text = remove_ansi.sub('', log_text)
@@ -955,7 +955,7 @@ def main():
             logfile, rc = process(source)
             errors += rc
             if not os.stat(logfile).st_size == 0:
-                with open(logfile, 'r') as f:
+                with open(logfile, 'r', encoding='utf8', errors='ignore') as f:
                     log_text = f.read()
                     if args.no_text_color:
                         log_text = remove_ansi.sub('', log_text)

[mauzey1]

@piotr-florek-mohc

Okay, I have added the suggested changes to PrePARE. I would still like to keep finding bugs in CMOR_CV code similar to the ones I have found, but this will at least reduce the chances of PrePARE crashing when processing erroneous NetCDF files.

[durack1]

@mauzey1 @piotr-florek-mohc (and @matthew-mizielinski) thanks for this tweak!

[mauzey1]

The MacOS Python 3.9 check failed due to a server error with conda. The last change committed would not have affected the tests anyway. I will merge this now.