All notable changes to this project will be documented in this file.
- Write an initial
status.csvas soon as the study graph has been constructed. Previously, the status file was not written until after scheduler submission of the first job step (when an asynchronous job step), or completion of the first job step (for a synchronous job step). Implemented in PR #477.
-
get_used_parametersno longer matches a parameter whose name is a prefix of another, so a step using only$(NP)is no longer reported as also using$(N). The token pattern now requires an exact name followed by an optional.attribute, and the name is escaped before it goes into the regex. Implemented in PR #478. -
Guards sleep in submit loop to only trigger if a submission failure occurs. Significantly speeds up job submission rates which were previously capped at 1 per second due to this sleep. Implemented in PR #481.
- Add what's new section to publish changelog and additional roadmap related material. Implemented in PR #472.
-
Creates ci profile for hypothesis enabled tests to account for increased system variability via extending deadlines. Implemented in PR #480.
- Added
autoyessupport to the cancel command. PR #461 - Added a compact, sortable hashing strategy for step IDs and workspaces. PR #465
- Added extra Flux options. PR #459
- Updated Flux exclusive behavior to modify task information passed to allocation and launcher logic. PR #467
- Made the Flux adapter aware of banks and queues. PR #453
- Enabled updating study configuration during execution. PR #445
- Dropped Python 3.7 and updated build system dependencies. PR #448
- Sanitized paths of serialized scripts in script adapters. PR #451
- Fixed a typo in
index.md. PR #439
- Added a
sacctfallback to the Slurm adapter to improve job tracking robustness. PR #405 - Added pager functionality to the
statuscommand. PR #420
- Synced the Read the Docs configuration with Poetry-based development environments. PR #399
- Updated Flux job state mappings for Flux versions
>= 0.26. PR #407 - Updated version information to be driven exclusively from
pyproject.toml, and connected it to the command line. PR #419 - Refactored the Flux adapter to avoid using
picklewhen communicating with Flux brokers installed in external environments. PR #415
- Printed command-line usage when no arguments are provided. PR #404
- Added a missing shebang in unscheduled scripts from the LSF adapter. PR #411
- Fixed Dockerfile issue
DL3006. PR #418 - Patched broken Flux job cancellation. PR #428
- Insulated Slurm adapters from user customization of
squeueandsacctoutput formats. PR #431
- Ported documentation to MkDocs and expanded feature and tutorial coverage. PR #403
- Bumped
certififrom2021.10.8to2022.12.7to address a security issue. PR #409 - Bumped
cryptographyfrom37.0.1to38.0.3to address a security issue. PR #410 - Updated the Poetry lockfile to address Dependabot-flagged security issues. PR #412
- Pinned Mermaid to
< 10.xdue to an API change. PR #422 - Bumped locked
certififrom2022.12.7to2023.7.22to address a security issue. PR #426
- Delivered extensive improvements to the Slurm, LSF, and Flux adapters for cluster scheduling.
- Added a Rich-enabled status view.
- Improved workspace hashing.
- Improved parameter generation.
- Included additional background improvements and bug fixes.
- Made the
Specificationclass easier to inherit. PR #280
- Fixed a regression affecting the use of variables and parameters in scheduling parameters for study steps. PR #279
- Updated the docs release version to
1.1.7. PR #278 - Added parameter generation documentation. PR #275
- Added the README as the long description so PyPI can render the Markdown README. PR #269
- Replaced a Unicode quote with an ASCII quote. PR #277