Open DavidSagan opened 6 years ago
Regarding directory and group, it is true that there is redundancy, and we should keep only one of them. Personally, I would prefer "group", since "directory" could lead to confusions with the file system's directories. Note also that the standard mentions "The used hierarchical data file format must provide the capability to create groups and sub-groups (in-file directories)", so from then on, it seems natural to use exclusively the word "group".
Regarding record and dataset: they are not exactly equivalent (although the way it is explained right now in the standard is confusing). For vector fields on a mesh (e.g. E), the record is E
and the datasets are the components E/x
, E/y
and E/z
. The reason we make this distinction is that some quantities (e.g. unitSI
) are more naturally stored at the record level, instead of the dataset level.
For scalar quantities on a mesh (e.g. rho
), record and dataset is the same thing, though.
Does this make sense?
See also here for an explanation of the record vs dataset: http://www.openpmd.org/#/step-9
Thanks for the explanation. I will use "group" in my draft.
the names "Directory", and "group" as far as I can tell, are used nearly interchangeably for the same thing. Should the Standard be simplified and only one name be used? Personally, I like "directory" but "group" would fine as well.
Also it looks like "record" and "data set" are used interchangeably. Should only one of these names be used in the Standard?
Thoughts?