Closed brancomat closed 3 years ago
What's happening is that arkimet sees a file, doesn't recognise the extension, and defaults to reading it as the result of an arki-query
(without --inline
or other switches).
I can see that the filename is config
and read it as a config file. However, arki-mergeconf also needs the name of the dataset, that is not in the config file.
Shall I say that if a path passed to arki-mergeconf is in the format .../$dirname/config
, then I parse config
as a configuration file, and add [$dirname]
and name=$dirname
to it?
We can change various arkimet tools not to fallback to some arbitrary format if the file name has no known extension, but to give an error instead.
So:
config
are parsed as config files with the dataset name being the containing directory nameOk for me and @edigiacomo
Let's say we have a dataset path populated with subdirectries with dataset config files.
This creates a working config:
The generated output looks like this:
This should be equivalent, but it isn't:
The generated output looks like this:
that results in an invalid config (
type=file
is meant for GRIB/BUFR files) and it's not coherent witharki-mergeconf
manpage (the config files should be valid sources).