cmd ref: freeze/unfreeze 1.0 and improvements

[jorgeorpinel]

Continuation of #1488

[jorgeorpinel]

Is this paragraph correct? Are import stages frozen by default, but they can be unfrozen so that repro re-downloads their outputs?

[shcheklein]

It's tricky and not very consistent (import vs import-url) - @efiop could you clarify how does it work these days? Also, we should start getting rid of the "import stage" terminology. At least while they stay out of dvc.yaml.

[jorgeorpinel]

Maybe @pared knows. I can try this later if not.

[pared]

Is this paragraph correct? Are import stages frozen by default, but they can be unfrozen so that repro re-downloads their outputs?

Seems to be true. Example:

#!/bin/bash

rm -rf repo other
mkdir repo other
main=$(pwd)

echo data >> data

pushd other

git init --quiet && dvc init --quiet
echo data >> data
dvc add data

git add -A
git commit -am "ble"
popd
pushd repo

git init --quiet && dvc init --quiet

dvc import-url $main/data imp-url
dvc import $main/other data -o imp 

popd
pushd other
echo modify >> data
dvc commit -f data.dvc
git commit -am "something"

popd
echo modify >> data

pushd repo
dvc status

dvc unfreeze imp-url.dvc
dvc unfreeze imp.dvc
dvc repro imp-url.dvc imp.dvc

It's tricky and not very consistent (import vs import-url) - @efiop could you clarify how does it work these days? Also, we should start getting rid of the "import stage" terminology. At least while they stay out of dvc.yaml.

the main difference between import-url and import results is how dvc status will react to them - in case of import-url we will get warning that it is frozen, in case of import-ed file the status will be checked and returned appropriately.

[jorgeorpinel]

Thanks @pared! So this paragraph is correct so the main point here is resolved. However we have these 2 side issues that may be worth exporting somewhere else:

we should start getting rid of the "import stage" terminology. At least while they stay out of dvc.yaml.

@shcheklein Idk. import .dvc files have both deps and outs so they're full stages. Added a checkbox about this to #608 for now.

import-url we will get warning that it is frozen, in case of import-ed file the status will be checked

@pared do you think both should behave the same though? Maybe both like regular import, just check the status, no warning. Or both warn + check status properly. If so maybe we should open a separate issue about this in the core repo.

[shcheklein]

@pared

the main difference between import-url and import results is how dvc status will react to them - in case of import-url we will get warning that it is frozen, in case of import-ed file the status will be checked and returned appropriately.

should it be a bug?

have both deps and outs so they're full stages.

deps and outs do not necessarily mean that something is a stage. It's an implementation detail. Probably a leftover for the backward compatibility. From a user perspective it's just an input data (similar to dvc add) to my mind. It's imported, yes, but it does not make any difference I think.

I agree that it is a separate discussion though and a separate ticket.

[jorgeorpinel]

It's an implementation detail... From a user perspective it's just an input data (similar to dvc add)... It's imported, yes, but it does not make any difference I think.

Yeah, you're right. We should get rid of the term then 🙂 (noted in #608)

[pared]

I will take on changing the messages in core

[jorgeorpinel]

@pared do you approve the changes now?

[pared]

@pared do you approve the changes now?
@jorgeorpinel, sorry for delay, I do