[Feature] Add `graphman unlink`

[lutter]

Description

Add a command that takes an IPFS hash and removes that deployment from all subgraph names that are using it as their current or pending version. In more detail: if a subgraph uses the hash as its current version, set current_version = pending_version and pending_version = null. If it uses it as the pending version, set pending_version = null.

This command should only accept an IPFS hash, not a subgraph name or sgd identifier.

Currently, this is done with the following pl/pgsql proc which also maintains an auditing table info.deleted_subgraphs; it would be good to keep this auditing, too. The code would have to create the table if it does not exist yet.

create or replace function unlink_deployment(did text)
returns void as
$$
declare
  sg record;
begin
  -- We only handle current and pending versions, even though the code
  -- talks about unuded versions for simplicity. We filter unused versions
  -- out in the queries.
  insert into info.deleted_subgraphs(id, name, sid, version,
                                     deployment, namespace, shard,
                                     subgraph_created_at, version_created_at)
  select ds.id,
         s.name,
         s.id as sid,
         case when v.id = s.current_version then 'current'
              when v.id = s.pending_version then 'pending'
              else 'unused' end as version,
         ds.subgraph as deployment,
         ds.name as namespace,
         ds.shard,
         to_timestamp(s.created_at) as subgraph_created_at,
         to_timestamp(v.created_at) as version_created_at
    from subgraphs.subgraph s,
         deployment_schemas ds,
         subgraphs.subgraph_version v
   where v.subgraph = s.id
     and v.id in (s.current_version, s.pending_version)
     and ds.subgraph = v.deployment
     and v.deployment = did;

  for sg in
    select s.id as sid,
           case when v.id = s.current_version then 'current'
              when v.id = s.pending_version then 'pending'
              else 'unused' end as version,
           case when v.id = s.current_version then 1
              when v.id = s.pending_version then 0
              else 2 end as sort_key
      from subgraphs.subgraph s, subgraphs.subgraph_version v,
           deployment_schemas ds
     where s.id = v.subgraph
       and v.id in (s.current_version, s.pending_version)
       and ds.subgraph = v.deployment
       and v.deployment = did
     order by sort_key
  loop
    if sg.version = 'current' then
      update subgraphs.subgraph
         set current_version = pending_version,
             pending_version = null
       where id = sg.sid;
     elsif sg.version = 'pending' then
      update subgraphs.subgraph
         set pending_version = null
       where id = sg.sid;
     end if;
  end loop;
end;
$$ language plpgsql;

Are you aware of any blockers that must be resolved before implementing this feature? If so, which? Link to any relevant GitHub issues.

No response

Some information to help us out

• [ ] Tick this box if you plan on implementing this feature yourself.
• [X] I have searched the issue tracker to make sure this issue is not a duplicate.

[SozinM]

Mind if i take this issue?

[zorancv]

Mind if i take this issue?

@SozinM sure thing! I'm rather busy with another issue and also on vacation right now.

[github-actions[bot]]

Looks like this issue has been open for 6 months with no activity. Is it still relevant? If not, please remember to close it.