Polish perun import and viewdiff #257
Conversation
The profile stats format now defines the ordering before the unit. Also, when parsing the stats and imported profiles, some leading and trailing whitespaces are now being stripped.
Profile metadata may be specified in the import CLI now. A proper internal representation of profile metadata has been implemented so that metadata may have a tooltip information associated with them.
Relative paths (to csv files, machine info file, profiles, ...) are now always prepended with the import-dir path, if provided. Absolute paths ignore the import-dir path and are kept as is.
Profile specification, stats and metadata now collapse / show both LHS and RHS at the same time.
|
|
||
| Multiple import entries may be specified; an import entry is either a profile entry | ||
|
|
||
| 'profile_path[,<exit code>[,<stat value>]+]' |
There was a problem hiding this comment.
Why this change? And can you maybe write an example how his is run? That's better readable then nested parenthesized specification.
There was a problem hiding this comment.
The parentheses represent optional parts of the import entries. This change was done so that (a) CSV and CLI specification of profiles to import is unified, i.e., it is now possible to specify exactly the same import parameters using both CSV files and CLI, and (b) at least exit code may be specified on CLI alongside the profiles to import. However, the interface is backwards-compatible and simple profile paths, e.g., import.stack.gz may still be specified. An example input containing both exit code and some stats values (corresponding to some specified stats headers defined in option --stats-headers) is, e.g., 'import.stack.gz,0,18511.379883,367'.
This PR addresses several issues with the current perun import and viewdiff modules. Among other things:
I think this issue resolves the last checkbox of #229 (please confirm @tfiedor), now that stats and metadata may be fully user-defined, including various aggregation and comparison functions, etc.