Crispr paml

[yfan2012]

Summary

New abstract methods added to Platform base class

• copy_file() — Copy a single file between projects
• get_current_user() — Return authenticated user profile as a standardized dict

---

_find_or_create_path() — SevenBridges

Fix: Root path '/' now returns None instead of trying to create a folder with an empty-string name. Changed from stripping the leading empty string to filtering out all empty strings from the split.

Unit tests (test_sevenbridges_platform.py):

• test_find_or_create_path_root_returns_none

Integration tests (test_find_or_create_path.py):

• test_root_path_returns_none
• test_creates_single_folder
• test_creates_nested_path
• test_existing_folder_is_reused
• test_partially_existing_path

---

upload_file() — SevenBridges

Fix: dest_folder='/' now correctly uploads to root. After stripping the trailing slash, an empty string skips _find_or_create_path and sets parent_folder_id = None. Also safely handles _find_or_create_path returning None.

Unit tests (test_sevenbridges_platform.py):

• test_upload_file_to_root — dest_folder=None → project root
• test_upload_file_to_subfolder — creates path, uploads into it
• test_upload_file_with_destination_filename — renames on upload
• test_upload_file_skips_existing_when_no_overwrite — returns existing id
• test_upload_file_overwrites_existing — uploads despite existing file
• test_upload_file_returns_none_on_failure — failed status → None
• test_upload_file_strips_trailing_slash — /inputs/ → /inputs
• test_upload_file_dest_folder_just_slash — '/' → root
• test_upload_file_extracts_basename — /home/user/file.txt → file.txt

Integration tests (test_upload_file_integration.py):

• test_upload_to_root
• test_upload_to_subfolder
• test_upload_to_nested_subfolder
• test_upload_with_destination_filename
• test_upload_no_overwrite_returns_existing_id
• test_upload_with_overwrite
• test_upload_to_root_via_slash

---

get_files() — Arvados bugfix

Fix: Filter logic was broken — independent if blocks caused duplicate results, and a misplaced else (attached to the suffix check, not the overall block) appended unfiltered files. Rewrote as skip-on-mismatch (continue) pattern. Additionally, it now returns the fully nested subdir path as opposed to just the collection name with the file name appended (correctly can return /inputs/run1/sample1/file1.fq.gz as opposed to the incorrect /inputs/file1.fq.gz previously).

---

copy_file() — All platforms

Platform	Behavior
SevenBridges	Creates target folder path if file_path provided, uses file.copy() or file.copy_to_folder()
Arvados	Parses source keep URI, finds/creates destination collection (named from first path component), copies via Collection.copy()
NGS360	Logs warning, returns file unchanged (WES doesn't support file copy)

---

_list_files_in_folder() — SevenBridges

Fix: When _find_or_create_path returns None (root path), falls back to self.api.files.query(project=project).all() instead of failing with AttributeError on None.id.

---

get_current_user() — All platforms

Platform	Implementation
SevenBridges	self.api.users.me() → maps to standard dict
Arvados	self.api.users().current().execute() → maps fields
NGS360	GET /api/v1/auth/me with Bearer token; parses UserPublic response directly (no "data" wrapper); safely handles null full_name

[yfan2012]

This if/else pairing causes all files to be matched as long as the 'suffix' is not specified, but another filter is. I think this is incorrect. See test_filters_work_on_both_methods() in test_get_files_fullpath.py

[yfan2012]

Regarding previous testing for the get_files() method with nested dir structure:

• This method previously did not have any unit tests - only integration tests requiring a live connection
• The integration tests only covered cases where a file was uploaded into a collection with no nested dir structure - this is why the bug was missed by the tests.
• Note: all the major launchers (RNA-seq, WES, WGS) primarily use get_file_id(), so they are unaffected by the bug. scRNA-Seq, gRNA, HIVE, etc use thier own versions of the PAML, so are also unaffected.

New changes:

• integration tests now check cases with nesting dir structure
• get_files_fullpath() removed, and its more correct behaviour that reports the full path is now incorporated into get_files()
• For SBG, copy_file() is now also updated to natively change any potential file_id inputs into file objects (following the same pattern of download_file() and export_file())

[golharam]

CollectionName shouldn't be mentioned here in base_platform.py. That's Arvados specific.

[yfan2012]

Good point, the example now given is the more realistic /inputs/run1/sample1/file1.fastq.gz

[yfan2012]

Tested RNA/WES/WGS pipeline launchers with these changes in Arvados and SBG using P-00000000-0001. All behaving normally.