dvc import: misleading warnings and outputs

[dmpetrov]

$ dvc -V
0.70.0+3de46d

One data file was updated in the imported repo. However, it was not a file that I'm trying to update.

$ dvc update data/data.xml.dvc
WARNING: DVC-file 'data/data.xml.dvc' changed.
WARNING: Stage 'data/data.xml.dvc' changed.
Importing 'data.xml (https://github.com/dmpetrov/dataset)' -> 'data/data.xml'
Output 'data/data.xml' didn't change. Skipping saving.

• UI issues 1: It should not be imported since there are no changes. Output: Importing 'data.xml (https://github.com/dmpetrov/dataset)' -> 'data/data.xml'
• UI issues 2: Warning duplication DVC-file 'data/data.xml.dvc' changed. and Stage 'data/data.xml.dvc' changed.

PS: Updated file was dir1/file2 which is corresponding to stage file dir1.dvc

EDITED:

• Issue 3 (not UI): It turned out that data/data.xml.dvc was actually updated after the update command - the repo checksum was changed (while checksum was not updated in the changed file dir1.dvc) with dvc-file checksum. Is it expected?

• Issue 4 (not UI): When I update a stage that was actually changed I still see this output ... didn't change message while the outputs were changed and even cache was updated. This happens even when I import a single artifact from a repo (I've check in another project). Code:

$ dvc update dir1.dvc
WARNING: DVC-file 'dir1.dvc' changed.
WARNING: Stage 'dir1.dvc' changed.
Importing 'dir1 (https://github.com/dmpetrov/dataset)' -> 'dir1'
Output 'dir1' didn't change. Skipping saving.

[pared]

Reproduction script:

#!/bin/bash

rm -rf repo remote_repo storage git_repo
mkdir repo remote_repo storage git_repo

maindir=$(pwd)
pushd remote_repo
git init >> /dev/null && dvc init -q

set -x
set -e
echo data >> data
echo data2 >> data2
dvc remote add -d str $maindir/storage
dvc add -q data data2
git add .dvc/config data.dvc .gitignore data2.dvc
git commit -m "add data"
dvc push -q

popd
pushd repo

git init >> /dev/null && dvc init -q
dvc import ../remote_repo data

popd
pushd remote_repo

#edit data2 only
echo data_new >> data2
dvc add data2
git add data2.dvc
git commit -m "modify data2"

dvc push

popd 
pushd repo
#data supposedly changed
dvc update data.dvc

[dberenbaum]

This is much improved since the issue was opened.

The output of the script @pared provided in the old version was:

WARNING: DVC-file 'data.dvc' changed.
WARNING: Stage 'data.dvc' changed.
Importing 'data (../remote_repo)' -> 'data'
Output 'data' didn't change. Skipping saving.

Currently, it is:

Importing 'data (../remote_repo)' -> 'data'

This looks like everything other than issue 3 above has been addressed. Not sure what expected behavior is in this case where the repo has been updated but without changes to the specified file. @pared @efiop Any thoughts on this?

[efiop]

Seems like point 3 is expected as that's the level of granularity we are dealing with in dvc update. Closing.