deps is inconsistent between dvc.yaml and .dvc files

[antonkulaga]

When I have something like:

    deps:
      - path: http://ftp.ensembl.org/pub/release-102/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
        etag: W/"5f8d62ed-34863818"
        cache: false

It works find in .dvc files, but when I add it to one of the stages, I get an error as it assumes deps are string only.

stages:
  prepare_genome:
    frozen: true
    deps:
      - path: http://ftp.ensembl.org/pub/release-102/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
        etag: W/"5f8d62ed-34863818"
        cache: false
    cmd: bin/prepare_genome.sh
    out:
      - data/gwas/homo_sapiens_ensembl_102/Homo_sapiens.GRCh38.dna.primary_assembly.fa

the error is:

 'dvc.yaml' format error: expected str @ data['stages']['prepare_genome']['deps'][0]

the dvc version is 1.11.10 (latest .deb file).
I suggest that "deps" should behave the same in .dvc and dvc.yaml as otherwise it creates a lot of confusion.

[skshetry]

@antonkulaga, the old .dvc file support for stage is going to be removed in the upcoming release next month.

.dvc files for stages were deprecated since the introduction of dvc.yaml file in 1.0, and are only supported for dvc add/import data files.

[antonkulaga]

@skshetry I was not talking about having stages in .dvc files, but having:

deps:
      - path: http://ftp.ensembl.org/pub/release-102/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.primary_assembly.fa.gz
        etag: W/"5f8d62ed-34863818"
        cache: false

inside of the dependencies in dvc.yaml So far only "str" is supported for deps in dvc.yaml, I believe it should also support etags for external dependencies

[skshetry]

@antonkulaga, the checksums are kept in lockfile. You just have to write the entry as follows and then dvc repro it later:

stages:
  download_file:
    cmd: curl https://github.com/api/repos/iterative/dvc -s > data.json
    deps:
    - https://github.com/api/repos/iterative/dvc
    metrics:
    - data.json

cache: false were never honoured for deps and is not required.

[antonkulaga]

@skshetry I think the problem is that in the docs you talk a lot about editing .dvc files and barely mention editing of dvc.lock
The other problem is that dvc import-url only allows creating .dvc files while many teams prefer to keep everything in dvc.yaml+dvc.lock combination. I am closing this issue in favor of a new issue about additing a "store in dvc.lock" flag in dvc import-url #5379