Documentation
¶
Overview ¶
Package targetname names the target a command acted on, so a command that changes something can say where it changed it ([ADR-0078](../../docs/adr/0078-the-cli-points-at-a-target.md) §4).
The failure the target model introduces is acting on the WRONG target — deploying to a cluster while believing you were deploying to the managed product, or the reverse. It cannot be prevented by design, because both are legitimate things to do and neither CLI can know which was meant. The only control left is making it immediately visible: the recoverable form of that mistake is noticing in the same breath, and the unrecoverable form is hearing about it later from somebody else.
Two properties follow from that and are the reason this is a package rather than a helper in one binary. First, the name shown is the one the person chose in the picker — the same string `burrow auth status` prints — because a server URL or an internal identifier is not what they would recognise. Second, it names what the command ACTUALLY reached: a recorded target is named only when that target is what decided the connection, and the raw kube context is named otherwise. Naming a target a command did not reach would manufacture the exact mistake this exists to catch.
Both `burrow` and `burrow-agent` render it, so the shape lives here rather than twice.
Index ¶
Constants ¶
const KindNone = "none"
KindNone is the kind reported when no configured target decided where the command went: the CLI followed the ambient kubeconfig, which ADR-0078 §1 preserves, or a per-invocation flag overrode the selection. It is deliberately not an empty string, so a reader of the JSON never has to distinguish "no target" from "this field was not written".
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Named ¶
type Named struct {
Name string `json:"name,omitempty"`
Kind string `json:"kind"`
Context string `json:"context,omitempty"`
Endpoint string `json:"endpoint,omitempty"`
Override bool `json:"override,omitempty"`
Detail string `json:"detail"`
}
Named is the target a command acted on. It is emitted as the `target` member of a mutating command's JSON result, so an agent composing a result can say where the change happened, and rendered by Clause for the human line.
Name is empty unless a configured target decided the connection; Context is the kube context that was actually reached; Endpoint is set only when a --control-plane URL bypassed the target model entirely. Detail is the sentence form, and for a configured target it is exactly what `burrow auth status` prints for that target, so the two can never drift into describing the same target differently.
func For ¶
func For(cfg *localconfig.Config, kubeContext string, override bool) Named
For names what an invocation acted on, given the kube context it connected to and whether a per-invocation --context overrode the selection.
A configured target is named only when it is a Kubernetes target whose context IS the one that was reached. That covers the ordinary case (the target decided the context) and deliberately excludes two others: a command that resolves through the kubeconfig without consulting the target at all, and a Burrow Cloud target, which has no kube context and cannot be what a kubeconfig connection reached. In both, naming the recorded target would assert something untrue about where the change landed.
An override wins over any recorded target, because ADR-0078 §4 keeps --context working as a per-invocation override and what the person overrode it TO is what was acted on.
func ForControlPlane ¶
ForControlPlane names a control plane addressed directly by URL with --control-plane. That flag bypasses the target model rather than selecting within it, so there is no picker name to show and the URL is the honest answer. A URL is not a credential; the token that goes with it is never rendered here or anywhere else.
Source Files