Fail when cluster commands are called without a proper workspace set #9
Conversation
| @@ -41,6 +43,10 @@ func migrate(ctx context.Context, cluster orb.OrbCluster, dbReset bool, orbs []s | |||
| err = errors.New("orbs and databases have to be of the same size") | |||
| return | |||
| } | |||
| if len(orbs) == 0 { | |||
There was a problem hiding this comment.
this change is a stowaway here, just was testing and thought it would be nice to cover this case even if remote.
| @@ -62,12 +62,8 @@ func LoadConfig(path string) (cfg *Config, err error) { | |||
| v.SetConfigFile(filepath.Join(path, "omnigres.yaml")) | |||
| err = v.ReadInConfig() | |||
| if err != nil { | |||
| if _, ok := err.(*os.PathError); ok { | |||
There was a problem hiding this comment.
I think this change may affect omnigres run as it can run files from a directory without a yaml file – so handling the error there might be needed.
This assumes we agree that omnigres run should be able to run yaml-less directories (by just throwing in a default config as the removed code did) – so we'll need to handle it there, in run.go – if the file is not found (error is of a certain type denoting that) then supply NewConfig()
There was a problem hiding this comment.
I was actually dropping that behavior (running on yaml-less dirs) for the sake of simplicity. I thought run would be analogous to start in that sense.
Should we keep this behavior for the run command?
orb/config.go
Outdated
| return | ||
| } | ||
| return | ||
| msg := fmt.Sprintf("omnigres.yaml not found in %s. Try setting a different workspace with -w or running omnigres init to create a new project config.", path) |
There was a problem hiding this comment.
For the idea above we'd need to type the error.
But outside of typing, there's fmt.Errorf that accomplishes errors.New(fmt.Sprintf())
diogob
force-pushed
the
require-config-when-running-cluster-cmds
branch
from
20fb47d to
ef710e8
Compare
without a proper configuration the current CLI proceeds with an empty config that yields behavior hard to understand for a newcomer. Solution: We can be more strict requiring an existing config to run such commands. This also allows for better error messages guiding the user during the initial project creation.
When a cluster command such as migrate or start is called without a proper configuration the current CLI proceeds with an empty config that yields behavior hard to understand for a newcomer.
Proposed solution
We can be more strict requiring an existing config to run such commands.
This also allows for better error messages guiding the user during the initial project creation.