get-started: cache: false explanation

[skshetry]

dvc.org/content/docs/start/experiments.md

Lines 51 to 52 in 2fb72f7

	> `cache: false` means that those file are small enough and versioned directly
	> with Git.

The cache: false is an intention to do what's explained, it's not what it means though.

[shcheklein]

@skshetry not sure I follow, could you elaborate please? (or even better suggest the way you think it should be).

[skshetry]

Something like this?

With cache: false, caching of the output is skipped by DVC and is usually done either if the file are small enough or needs to be versioned directly with git.

[shcheklein]

ah, I see ... to be honest, I think it's fine to keep it simple, in 99.9% cases this is cache false means that those files are indeed tracked directly with Git. I haven't see any other cases yet at least.

We can do something like "In this case cache: false means that those files are small enough to be versioned with Git".

[skshetry]

Still, cache: false does not mean that those files are small enough or is versioned by Git. Probably, it's one of the usecase of it, but it is not what it means.

[shcheklein]

Still, cache: false does not mean that those files are small enough or is versioned by Git. Probably, it's one of the usecase of it, but it is not what it means.

yep, still in this case it's totally fine since we talk about metrics and plots file specifically - Get started is not command reference - no need to be mathematically precise.

[skshetry]

@shcheklein, it's under "Expand to see what happens under the hood", so, we can explain that detail, or we could rephrase it as,

With cache: false, DVC skips caching the output as we want "metrics.json" to be versioned by Git.

Also, I see that Data Pipelines explains all the flags, but "Experiments" does not explain what -M and --plots-no-cache is, and why -M is being used instead of -m and --plots-no-cache instead of --plots.

[shcheklein]

@skshetry I don't have a strong opinion on this - to my mind both ways (the current one and the one you suggested) are fine for this section and the meaning more or less the same. So, feel free to make a PR. The essence is the same - we version files directly with Git since, and we do this usually when they are small enough. I would still try to keep language simple and mention that files are small though.

Also, I see that Data Pipelines explains all the flags, but "Experiments" does not explain what

I would say it's a bit of a legacy that we explain flags in such details. But good point, and we can add that info if it's hidden.

[utkarshsingh99]

If you are going ahead with this change, can I take this up?

[shcheklein]

@utkarshsingh99 absolutely! feel free to create a PR.