Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

run: Use 'datalad add' in failure instructions #3080

Merged
merged 5 commits into from Dec 10, 2018

Conversation

@kyleam
Copy link
Contributor

@kyleam kyleam commented Dec 8, 2018

  • Minor improvments saves --message-file handling

  • Update add to support --message-file

  • Update runs failure instructions to use datalad add (closes #3072)

kyleam added 5 commits Dec 8, 2018
Under Python 2, we pass a bytestring to commit(), but the intention
was to pass a unicode string, like we do under Python 3.  Passing a
bytestring as the message doesn't seem to cause any issues, at least
according to test_save_message_file, but we should still do so for
consistency across Python versions.
When save is called from the command line with both --message and
--message-file but without -d<dataset>, we get two error messages, the
intentional error message as well as one about a key error:

  [ERROR  ] Both a message and message file were specified
  [ERROR  ] 'path' [utils.py:default_result_renderer:503] (KeyError)

This is because we pass refds_path as the path to get_status_dict(),
but refds_path is None in the above scenario.

Instead of worrying about properly populating the status dict info,
which isn't particularly relevant for the --message/--message-file
conflict, just raise a plain exception.  Doing so is in line with
what's done in other interfaces (e.g., aggregate raises a ValueError
when it receives an unknown update mode).
This will be used in `add` as well.
`save` learned to accept a --message-file argument in da8d235 (ENH:
save: Add -F/--message-file option, 2018-03-28).  Teach `add` to do
the same.
If a run command has a non-zero exit, we don't save the outputs and
instead instruct the user how to save the outputs given the failure is
expected.  The instructions say to use 'datalad save -r .', but there
are cases where this is not sufficient to add an output file (e.g., if
a file is added to subdataset).  Update the instructions to use 'add'
instead of 'save', which is what run actually uses (by default)
underneath.

Also add '-d .' to the instructions to match run's dataset.add(),
reducing the possibility of discrepancies like the one that cropped up
in dataladgh-2276.  There are still cases where run's dataset.add() is not
sufficient (a known one is dataladgh-2454, which is addressed in
datalad-revolution), but at least the command in the instructions will
match run's behavior.

Closes datalad#3072.
@kyleam kyleam force-pushed the run-fail-msg-add branch from ab6bd6a to 67ce0e3 Dec 8, 2018
@codecov
Copy link

@codecov codecov bot commented Dec 8, 2018

Codecov Report

Merging #3080 into master will increase coverage by <.01%.
The diff coverage is 100%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master    #3080      +/-   ##
==========================================
+ Coverage   90.24%   90.24%   +<.01%     
==========================================
  Files         248      248              
  Lines       32571    32589      +18     
==========================================
+ Hits        29393    29411      +18     
  Misses       3178     3178
Impacted Files Coverage Δ
datalad/interface/run.py 100% <ø> (ø) ⬆️
datalad/interface/tests/test_save.py 100% <100%> (ø) ⬆️
datalad/interface/tests/test_run.py 99.83% <100%> (ø) ⬆️
datalad/distribution/tests/test_add.py 82.95% <100%> (+0.53%) ⬆️
datalad/distribution/add.py 96.25% <100%> (+0.14%) ⬆️
datalad/interface/save.py 88.88% <100%> (+0.07%) ⬆️
datalad/interface/common_opts.py 100% <100%> (ø) ⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update de68577...67ce0e3. Read the comment docs.

@@ -611,7 +611,7 @@ def run_command(cmd, dataset=None, inputs=None, outputs=None, expand=None,
ofh.write(assure_bytes(msg))
lgr.info("The command had a non-zero exit code. "
"If this is expected, you can save the changes with "
"'datalad save -r -F %s .'",
"'datalad add -d . -r -F %s .'",
Copy link
Member

@yarikoptic yarikoptic Dec 9, 2018

Now thinking about it, this wouldn't be a correct instruction in case of --explicit run... Not sure what would be a sensible cross platform and concise way though, besides add (may be adding it as a feature for all datalad API functions getting a list of files) getting an option of taking a file with filenames to add, and then is preparing that file alongside with the message file to be given here.
I will leave it for you to decide to move that into a new issue/PR and here for explicit mode may be just say that it might add too much or listing all outputs?

Copy link
Contributor Author

@kyleam kyleam Dec 9, 2018

this wouldn't be a correct instruction in case of --explicit run

When I've thought about it in the past, I've decided it's not worth worrying about --explicit here. If you feel strongly that we should tailor the message, please open a separate PR tweaking the message as you see fit.

@yarikoptic yarikoptic merged commit c47a37a into datalad:master Dec 10, 2018
6 of 8 checks passed
@kyleam kyleam deleted the run-fail-msg-add branch Dec 10, 2018
yarikoptic added a commit that referenced this issue Feb 8, 2019
 A variety of bugfixes and enhancements

 ### Major refactoring and deprecations

 - All extracted metadata is now placed under git-annex by default.
   Previously files smaller than 20 kb were stored in git. ([#3109])
 - TODO: get_runner #3104 and pending #3131

 ### Fixes

 - Improved handling of long commands:
   - The code that inspected `SC_ARG_MAX` didn't check that the
     reported value was a sensible, positive number. ([#3025])
   - More commands that invoke `git` and `git-annex` with file
     arguments learned to split up the command calls when it is likely
     that the command would fail due to exceeding the maximum supported
     length. ([#3138])
 - The `setup_yoda_dataset` procedure created a malformed
   .gitattributes line. ([#3057])
 - [download-url] unnecessarily tried to infer the dataset when
   `--no-save` was given. ([#3029])
 - [rerun] aborted too late and with a confusing message when a ref
   specified via `--onto` didn't exist. ([#3019])
 - [run]:
   - `run` didn't preserve the current directory prefix ("./") on
      inputs and outputs, which is problematic if the caller relies on
      this representation when formatting the command. ([#3037])
   - Fixed a number of unicode py2-compatibility issues. ([#3035]) ([#3046])
   - To proceed with a failed command, the user was confusingly
     instructed to use `save` instead of `add` even though `run` uses
     `add` underneath. ([#3080])
 - Fixed a case where the helper class for checking external modules
   incorrectly reported a module as unknown. ([#3051])
 - [add-archive-content] mishandled the archive path when the leading
   path contained a symlink. ([#3058])
 - Following denied access, the credential code failed to consider a
   scenario, leading to a type error rather than an appropriate error
   message. ([#3091])
 - Some tests failed when executed from a `git worktree` checkout of the
   source repository. ([#3129])
 - During metadata extraction, batched annex processes weren't properly
   terminated, leading to issues on Windows. ([#3137])
 - [add] incorrectly handled an "invalid repository" exception when
   trying to add a submodule. ([#3141])
 - Pass `GIT_SSH_VARIANT=ssh` to git processes to be able to specify
   alternative ports in SSH urls

 ### Enhancements and new features

 - [search] learned to suggest closely matching keys if there are no
   hits. ([#3089])
 - [create-sibling] gained a `--group` option so that the caller can
   specify the file system group for the repository. ([#3098])
 - Interface classes can now override the default renderer for
   summarizing results. ([#3061])
 - [run]:
   - `--input` and `--output` can now be shortened to `-i` and `-o`.
     ([#3066])
   - Placeholders such as "{inputs}" are now expanded in the command
     that is shown in the commit message subject. ([#3065])
   - `interface.run.run_command` gained an `extra_inputs` argument so
     that wrappers like [datalad-container] can specify additional inputs
     that aren't considered when formatting the command string. ([#3038])
   - "--" can now be used to separate options for `run` and those for
     the command in ambiguous cases. ([#3119])
 - The utilities `create_tree` and `ok_file_has_content` now support
   ".gz" files. ([#3049])
 - The Singularity container for 0.11.1 now uses [nd_freeze] to make
   its builds reproducible.
 - A [publications] page has been added to the documentation. ([#3099])
 - `GitRepo.set_gitattributes` now accepts a `mode` argument that
   controls whether the .gitattributes file is appended to (default) or
   overwritten. ([#3115])
 - `datalad --help` now avoids using `man` so that the list of
   subcommands is shown.  ([#3124])

* tag '0.11.2': (124 commits)
  Changelog entry for GIT_SSH_VARIANT change
  ENH: Declare our GIT_SSH_COMMAND as GIT_SSH_VARIANT=ssh
  BF: sshconnector: Don't use ssh's port flag as scp's
  RF: sshconnector: Simplify shlex quote import
  CHANGELOG(0.11.2): Fix some typos
  [DATALAD RUNCMD] CHANGELOG: Linkify 0.11.2 entries
  CHANGELOG: Do first pass for 0.11.2
  CHANGELOG: Add missing link target for download-url
  Start cooking the 0.11.2 release
  RF: appveyor - move test_install tests to be ran the last
  RF: text_type instead of str
  ENH(TST): provide my timing for the slow test
  BF(TST): adjust the test for the fact that AnnexRepo.add does not blow on nonexisting files
  ENH(TST): two tests which test quick or thorough for add failing with too long list of files
  BF: get stderr if present, otherwise just use str(e)
  BF: append out/err only if not empty/None
  Centrlize handling running commands with long files list in _run_command_files_split
  RF: remove minor duplication of -- handling, place all files handling closer to the call
  RF: move unrelated to try/except handling outside
  ENH+DOC: Report actual process handle, not just PID
  ...
@yarikoptic yarikoptic added this to the Release 0.11.2 milestone Feb 10, 2019
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Linked issues

Successfully merging this pull request may close these issues.

2 participants