Re: [BuildStream] bst artifact show

[Tristan Van Berkom <tristan vanberkom codethink co uk>]

On Fri, 2019-05-31 at 16:28 +0000, Benjamin Schubert wrote:
> Hey everyone,

Hi Ben,

I think technically we are revisiting some aspects of:

    https://mail.gnome.org/archives/buildstream-list/2018-July/msg00051.html

Lets remember to take these previous threads into account :)

I did not raise `bst artifact show` in that initial proposal but over
time I can recall presuming that such a command which should have
semantical symmetry with `bst show` for elements should definitely
exist.

I certainly think this should be scriptable like `bst show` and allow
specifying format strings which is convenient for scripts which parse
the stdout, and should support the same wildcarding for artifact refs
which is intended to be supported for the `bst artifact` commands.

I also think that following the previous discussions of `bst artifact`
command subgroup, the `bst artifact show` should probably *replace* the
notion of `bst artifact list`, as it would be more symmetrical with
`bst show` and would allow one to cover all of the same use cases.

I'm particularly interested in being able to list creation dates of
artifacts, list files within an artifact, overall size of artifact and
sizes / filesystem attributes of individual files listed in an
artifact.

> As part of a system that we are building, we would like to be able to
> validate that an element can be pulled from a remote cache.
> Therefore I would like to propose a new `bst artifact` subcommand:
> `bst artifact show`.
>
> This command should be able to show the state of an artifact on cache servers.

I share James's concern in the adjacent reply here.

I would certainly expect by default to be working with the local
artifact cache, which I have control over, and I can predict that
subsequent results of `bst artifact show` will be consistent unless I
have run any local commands which might result in deletion of the
artifact (i.e. `bst artifact delete`, or commands which might trigger a
cleanup) - I also think this expectation is important for any scripting
purposes.

That said, of course it would make sense to also have the ability to
query a remote artifact server with this command, I just think this
should be an explicit option on the `bst artifact show` command line
(probably it should *also* support specifying which particular remote
url should be queried, overriding user configuration and hints from
project data, as we do in other commands like `bst push` and `bst
pull`).

Cheers,
    -Tristan