Normalize paths in packet metadata. #46
Conversation
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #46 +/- ##
==========================================
- Coverage 99.28% 98.70% -0.58%
==========================================
Files 45 45
Lines 3351 3480 +129
Branches 527 337 -190
==========================================
+ Hits 3327 3435 +108
- Misses 21 39 +18
- Partials 3 6 +3 ☔ View full report in Codecov by Sentry. |
plietar
force-pushed
the
mrc-5241
branch
3 times, most recently
from
508266a
to
507ca51
Compare
| @@ -200,3 +200,30 @@ def openable_temporary_file(*, mode: str = "w+b", dir: Optional[str] = None): | |||
| os.unlink(f.name) | |||
| except OSError: | |||
| pass | |||
|
|
|||
|
|
|||
| @overload | |||
There was a problem hiding this comment.
Yeah, it makes it possible to describe that the output type depends on the input. Might be overkill in this context
| join("shared_data", "numbers.txt"): join("data", "numbers.txt"), | ||
| join("shared_data", "weights.txt"): join("data", "weights.txt"), | ||
| "shared_data/numbers.txt": "data/numbers.txt", | ||
| "shared_data/weights.txt": "data/weights.txt", |
There was a problem hiding this comment.
is it worth adding something that checks (perhaps on cleanup of a temporary orderly directory) that we never get backslashes anywhere?
There was a problem hiding this comment.
How about we do that as part of the JSON schema? We can have a regex on the path-like fields.
Wouldn't be the friendliest of error messages though.
There was a problem hiding this comment.
This is a great idea, because it will get caught in the implementations at least
The packet metadata keeps a record of what files are used and/or produced by the report. When such files are directories, we want to normalize the path separator. This will ensure that a packet produced on Windows can be used on POSIX platforms and vice-versa. Additionally, R (and therefore orderly2) uses forward slashes on all platforms, including Windows. Using forward slashes in outpack-py means packets are compatible and consistent across the two tools. It's not obvious whether the functions that return copied paths to the caller (eg. `orderly.shared_resource`) should return native paths or POSIX paths. For now I've kept these as native paths.
The packet metadata keeps a record of what files are used and/or produced by the report. When such files are directories, we want to normalize the path separator. This will ensure that a packet produced on Windows can be used on POSIX platforms and vice-versa.
Additionally, R (and therefore orderly2) uses forward slashes on all platforms, including Windows. Using forward slashes in outpack-py means packets are compatible and consistent across the two tools.
It's not obvious whether the functions that return copied paths to the caller (eg.
orderly.shared_resource) should return native paths or POSIX paths. For now I've kept these as native paths.