* use setuptools_scm and build_requires
* remove kwargs from setup.py
* fix tests; fetch in depth in github actions
* use build to build sdist/wheel
* try updating pip and setuptools
* update pip
* remove or unpin wheel
* bring back the version in pyinstaller
* Update scripts/build.py
Co-authored-by: Ruslan Kuprieiev <[email protected]>
* advertise Python 3.10 support
Pygit2 does not have windows wheels yet and no pyarrow
for 3.10. But these platform-dependent things and
remote-dependent things should not stop us from
advertising support as we are pure-python based
package and we cannot claim support for all of our
dependencies.
* set pygit2 for Python 3.6 to <1.7
Dependabot was throwing Illformed requirements `"< '3.10'"` for the
hdfs. But I noticed that it did accept other requirements where we
do similar. So it seems the indentation fixes it.
Even then, dependabot did not update any dependencies. I removed
setup.py file, after which immediately it started creating PRs.
So, I recknoed that it must have been due to some mechanism
in place from dependabot to disallow execution, so I tried to
remove few blocks of code from setup.py. And, I was eventually
able to figure out that it was from the block where we import version.
So, I added a try-except block and it magically again worked. :)
Let's see how it goes.
* exp init: introduce `--with={default,live}` flag
This flag will select the different mode of outputs of stage.
With `--with=live`, it'll enable `live` setting in the stage.
On interactive mode, it'll always ask user if they want to
use live (unless `--with=live` is specified), and then it'll
skip metrics/plots question and moves on to ask for path to
dvclive path.
Also, if live setting is enabled in the stage, it'll set the
`models` to be a checkpoint output.
* s/with/type and don't ask for live interactively
This was used before on opening http urls as a stream,
but we no longer use those (as we use fsspec's open method).
One thing to note however is that, compared to previous
implementation, fsspec's http filesystem's open is not very
robust to failures.
* tests: additional callbacks test for remaining filesystems
* requirements: bump pydrive2 to 1.9.4
Co-authored-by: Ruslan Kuprieiev <[email protected]>
Trying to see if getting rid of this dependency, our issue with timeout
gets solved or not.
We use timeout-minutes anyway on GHA, so we probably don't need it.
Locally, users can monitor this themselves.
* fs.download: add callback support
* fix lint issues
* compat for gdrive, use fs to upload text for tests
* fix typing issues
* skip gdrive test
* Update tests/func/test_fs.py
* fs.upload: add support for optional callback
Also removes _upload methods in other fses, and replaces them
with fsspec-compatible put_file which supports fsspec callback
even if they are not fsspec based filesystem.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
* Set timeout for jobs in Github Actions
Sometimes tests are getting frozen(most likely an issue with `pytest-xdist`), and sometimes the pip resolver is slow due to some weird constraints set by us or one of our dependencies.
The default timeout is 360 minutes (6hrs), but our test completes on ~18 minutes at maximum, so I think 30mins should be enough to account for occasional CI slowdown.
Also keeping the total job timeout at 45 mins, we usually require just 4 mins of extra time to do installation and setup but giving it 11 more minutes just in case.
Similarly, `lint` completes in <5 mins usually, but giving it a 10 mins timeout.
This should give us early feedback on the issue.
* Update setup.cfg
* Update setup.cfg
* extract transfer related logic from odb to fs.utils.transfer
* don't makedirs on dirinfo
As local odb needs to set a mode for dir, we cannot create a directory
in the transfer
Making sure that script takes a bit more time than the await time
I am not really sure this was the cause though.
Also the tests are flaky only in MacOS, but I haven't been able to reproduce.
* add support for Python 3.10
pyarrow is not available, so we are skipping tests for it.
And, restricting `hdfs` extras to be not installed in Python3.10.
Pathlib removed `_init` method and `init` parameter, so just working
around that. It's not a big change.
* skip windows on 3.10; update pip/wheel/setuptools during install
pywin32 is not yet available for 3.10 Python, so skipping those for now
* install libgit2 with Python3.10 as pygit2 wheel is not available
* install libgit2
* only run on macOS, libgit2-1.1 not available on ubuntu
* demonstrate issue while sending analytics report
When sending analytics, we are not flushing the data to the temp file
which causes failure at times, especially on Windows
* close file before calling daemonized analytics send command
This way, the contents bufferred will get a chance to be flushed.
* Fix merging of params when empty
Recent release of python-benedict==0.24.1 started ignoring empty dictionaries
during merge. The way we expected to work was questionable and also how
benedict itself tries to handle the references to empty dictionary is
also questionable.
I consider this to be a proper way to handle this even if this gets fixed
on the benedict side.
* Revert "setup: limit python-benedict version (#6377)"
This reverts commit 15de826266.
* add comment for why passing through benedict
* add: do not delete stage files before add
Because of the way we collect stages and cache them, we were not able to
collect them for the `add` without removing them from the workspace.
As doing so, we'd have two same/similar stages - one collected from the
workspace and the other just created from the `dvc add` in-memory.
This would raise errors during graph checks, so we started to delete them
and reset them (which is very recently, see #2886 and #3349).
By deleting the file before we even do any checks, we are making DVC
fragile, and results in data loss for the users with even simple mistakes.
This should make it more reliable and robust.
And, recently, we have started to keep state of a lot of things, that by
resetting them on each stage, we waste a lot of performance, especially
on gitignores. We cache the dulwich's IgnoreManager, which when resetted
too many times, will waste a lot of our time just collecting them again
next time (see #6227).
It's hard to say how much this improves, as this very much depends on
no. of gitignores in the repo (which can be assumed to be quite in number
for a dvc repo) and the amount of files that we are adding
(eg: `-R` adding a large directory).
On a directory with 10,000 files (in a datadet-registry repo),
creating stages on `dvc add -R` went from 64 files/sec to 1.1k files/sec.
* add tests
* make the test more specific
* separate validation and transformation of args from add func
* make tests pass for Python3.6
* cleanup
* use dict for attr-access and patch boundarguments.arguments
astroid==2.5.7 has improved inference and was throwing
errors that were false negatives before. Fixing these
and ignoring some.
Also added `funcy.decorator` to the `signature-mutators`,
as it changes function signature (i.e. provides `Call` as
a first argument which is not required to pass).
Example errors that we were getting because of this:
```
tests/remotes/gdrive.py:40:11: E1120: No value for argument 'tries' in function call (no-value-for-parameter)
************* Module dvc.stage.serialize
dvc/stage/serialize.py:38:1: E1120: No value for argument 'func' in
function call (no-value-for-parameter)
************* Module dvc.fs.http
dvc/fs/http.py:19:1: E1120: No value for argument 'ctx' in function call (no-value-for-parameter)
************* Module dvc.fs.gdrive
dvc/fs/gdrive.py:66:11: E1120: No value for argument 'tries' in function call (no-value-for-parameter)
dvc/fs/gdrive.py:459:5: E1120: No value for argument 'ctx' in function call (no-value-for-parameter)
```
Saugat Pachhai
