iterative/dvc 2.0.0a0

2.0.0a0 🦉

on GitHub

Breaking change

• Changed dvc.lock schema.
  It is still backward compatible in terms of read and
  will be migrated to the new format whenver it's written (#5222).

Deprecations and Removals

• Removed --cwd from dvc repro (#5048).
• Deprecated dvc run command. dvc stage add is now recommended instead.

New Features

• dvc.yaml supports parametrization and looping constructs.
• Support list of commands in cmd field in dvc.yaml file (#4934).
• Added experiments feature and dvc exp command.
• Added support for straight-to-remote feature.
• Added support for live, and corresponding dvc live command.
• Added stage:add command as a replacement for dvc run.
• Added stage:list command to list all the stages in the repo.
• Added --glob for wildcard pattern as target option for add/push/pull (#4864, #5033, #5032)
• diff/params diff/metrics diff got --targets flag (#4938, #5122)
• config: Support listing config through --list and their source with --show-origin (#5080, #5188)
• Added '--show-md' flag for 'dvc metrics show' (#4770) (#5161)

Bug Fixes

• dvcignore: Fix incorrect ignored output computation (#4986)
• fix partial file checkouts in the same directory (#5163)
• diff accounts for deleted workspace data (#5205)
• fix --help output (#5193, #4999)
• config: read merged config by default (#5179)
• aliyun-oss: fix pull error for large file (#5053)
• Run graph checks on collect/find_outs_by_path (#5035)
• follow the symlink when deciding filesystem type on dvc version (#5067)
• Fix get-url failing with AttributeError (#5118)

Enhancements

• .dvc files can store executable permissions (#5061).
• dvc metrics displays as a table (#4706).
• metrics/params diff/show table are now ordered.
• commit: support granularity (#4297) (#5088).
• dvc import can be passed --jobs (#4977).
• diff can detect file renames (#5160).
• Added support for autocompletions of stage names in zsh/bash.
• Add ">" to script during dvc run/repro (#5041)
• Reducing clutter of log messages (#5017, #5236)
• Added dvc repo was initialized when dvc init-ed (#5190)
• make IterStream seekable and peekable (#5084)
• Improve OverlappingOutputsError exception message. (#5015)
• delete .gitignore file if empty (#5148)

Optimizations and Performance

• push: Gradual task allocation for pushing big datasets (#5285)
• Startup time has been improved.
• git: don't list all commits to check for an empty repo (#5075)

Internals and Libraries

• Replace deprecated pyarrow.hdfs.connect (#4973)
• psutil is now a requirement (#5232)
• Increment shtab's version (#5212)