Module projection

Expand description

The one pure target-specific projection over this binary’s embedded sources.

Projection::compute takes a ProjectionInput, values alone, and answers the complete candidate artifact tree this installed binary would land in one target: every destination with its ownership Kind, its Placement, the complete proposed bytes, the rendered region where the destination is a marked region, and the embedded source paths it was rendered from. Staging and production landing share it byte for byte, so what an agent studies in a stage is what a landing writes.

This module reads snippets and blocks through src/embedded.rs directly and reads nothing else: no target file, no environment, no Git state, no clock, no registry, no network. Whatever a projection needs from the target arrives as TargetEvidence, gathered before construction by evidence::gather, which lives in its own file so a source scan can hold this one to the pure boundary.

Every pure piece of the landing model has one implementation here: the kind table, the token substitution, the block templating, the splice and marker judgments, and the Nix crate-shape judgment. Which embedded source belongs to which capability, and whether a capability is available at the target’s dimensions, is the catalog’s answer in crate::profile::catalog, and this module renders what the catalog selects. src/landing.rs re-exports the pieces under their old names.

One input type, one compute function, one candidate shape: staging and every landing verb consume this one projection.

Re-exports§

pub use crate::landing::manifest::Provider;

Modules§

evidence: Gathering the target evidence a projection consumes.

Structs§

Candidate: One proposed destination.
Collision: One block destination the target’s document cannot take.
CrateShape: The structural facts of the target’s crate that the seeded Nix package expression and the seed flake’s smoke check rely on.
Omission: One destination withheld from this target, with why.
Projection: The complete candidate artifact tree for one target.
ProjectionInput: The complete input to one projection: the resolved landing parameters and the typed evidence read from the target beforehand.
TargetEvidence: What a projection needs to know about the target, gathered before the projection runs and carried as values.
TrunkGuards: The two authored guard entries the integration mode selects between, beside the mode itself.

Enums§

Kind: Who owns a landed file’s bytes after landing.
Placement: How a candidate occupies its destination.

Constants§

AGENTS_BLOCK: The authored routing-block template.
AGENTS_DESTINATION: The destination the routing block splices into.
AGENTS_LINE_BRANCHES: The routing block’s mode line, branches form.
AGENTS_LINE_INTEGRATION_FORGE: The routing block’s integration line, forge form.
AGENTS_LINE_INTEGRATION_LOCAL: The routing block’s integration line, local form.
AGENTS_LINE_WORKTREE: The routing block’s mode line, worktree form.
BLOCK_BEGIN: The block’s opening marker.
BLOCK_DESTINATIONS: Every block destination, in the order a landing writes them.
BLOCK_END: The block’s closing marker.
BRANCH_GRAMMAR: The one branch grammar.
BRANCH_SHAPE_TOKEN: The head-branch shape the release driver gives its own request.
CODE_SCANNING_DESTINATIONS: Every destination the opt-in code scanning capability can land, against the provider that lands it.
CODE_SCANNING_TECHS: The bindings that ship a code scanning workflow.
GITLAB_ROOT_PIPELINE: The GitLab root pipeline, the one destination a release-less landing withholds where the target owns it.
GITLAB_TITLE_FRAGMENT: The GitLab title fragment the root pipeline includes.
GLOSSARY_BLOCK: The authored glossary template.
GLOSSARY_DESTINATION: The destination the glossary block splices into.
HOOKS_BEGIN: The hook block’s opening marker, a YAML comment at column zero.
HOOKS_DESTINATION: The destination the hook block splices into.
HOOKS_END: The hook block’s closing marker.
HOOK_TYPES_LINE: The top-level key the fresh hook file carries and the skills verify on an existing one: the commit-msg and pre-push hooks run only where their hook types are installed.
LINE_PREFIX_RE_TOKEN: The same prefix, escaped for a slash-delimited regular expression.
LINE_PREFIX_TOKEN: The release-line branch prefix, naming the lines a release trigger accepts beside the trunk.
NIX_DESTINATIONS: The destinations of the opt-in Nix capability, present in a projection only where the landing’s nix parameter is on.
NIX_WITHHOLDABLE: The subset a target with a flake of its own keeps out: the seed pair, whose files would sit beside a flake release-kit did not author.
OWNER_TOKEN: The mechanical substitution sites in rendered files.
PRE_COMMIT_BLOCK: The authored hook-block template.
PRE_COMMIT_INTEGRATE_NOTE: The header note naming the pre-integrate contract.
PRE_COMMIT_SWEEP_NONE: The header note for the pairing whose sweep skips nothing.
PRE_COMMIT_SWEEP_SKIP: The header note naming what a CI sweep skips.
PRE_COMMIT_TRUNK_COMMIT_GUARD: The forge integration mode’s guard against a trunk commit.
PRE_COMMIT_TRUNK_PUSH_GUARD: The forge integration mode’s guard against a trunk push.
PRE_COMMIT_WORKTREE_GUARD: The worktree mode’s guard entry.
PUSH_GUARD_TOKEN: The token a release workflow’s job condition carries, so that a rendering which adds a second trigger cannot let the new event enter a job written for a push.
RELEASE_GATE_JOB: The gate’s own job, added to a release workflow.
RELEASE_GATE_TRIGGER: The gate’s trigger block, added to a release workflow’s on key.
REPO_PLACEHOLDER: The repository a preview stands in for where nothing answered.
REPO_TOKEN: The full recorded project path, including nested namespaces.
REQUEST_JOB_TOKEN: The id of the job that opens and refreshes the release request.
REQUIRED_CHECK_TOKEN: The check the release gate judges.
REQUIRED_WORKFLOW_TOKEN: The workflow whose completion wakes the release gate.
SCOPE_SHAPE: The one commit scope shape.
SCOPE_SHAPE_TOKEN: The one scope shape: the title checks’ regular expression.
SCORECARD_DESTINATIONS: The destinations of the opt-in Scorecard capability, present in a projection only where the landing’s scorecard parameter is on.
SECURITY_SPANS: The three replaceable spans of a landed security policy, each as its ordered begin and end marker.
STYLE_TOKEN: The recorded release style: trunk arms the bot’s request in the landed release workflow, lines leaves every request unarmed.
TRUNK_BRANCH_TOKEN: The one permanent branch. A landed release trigger, ref guard, and branch guard each name it, so a target whose trunk is not master needs its own answer in its own bytes.

Functions§

authored: An authored block without the one final newline the repository’s hooks enforce on every file under blocks/; a test in src/embedded.rs holds each file to exactly one.
block_markers: The markers of a block destination, or None for a whole-file one.
code_scanning_incompatibility: Why the named code scanning provider cannot run at this (technology, forge), or None where it can or none is named.
code_scanning_licence_refusal: Why the code scanning capability’s licence condition refuses this target, or None where no condition applies or the licence satisfies it.
compose_glossary: The glossary block from its authored template: markers included and without a trailing newline. It carries no token and no mode, so the same bytes land in every target.
compose_hooks: The hook block from its authored template, for one pairing of the two Git workflow axes.
compose_routing: The routing block from its authored template and the mode’s one orientation line.
destinations: Every destination the embedded sources can land, in declaration order.
embedded_block: One authored block this binary embeds, as text, by its embedded path.
extract_block: The marked block inside a document, markers included, or None where the text carries no complete block.
flake_pair_withheld: Why the flake half of the Nix capability stays out of this landing, or None where the pair lands whole.
glossary_block: The glossary block from this binary’s embedded template.
hooks_block: The hook block for one checkout mode, from this binary’s embedded templates.
hooks_marker_defect: marker_defect for the hook file, the destination whose entries execute.
integration_line: The routing block’s integration line for one integration mode.
kind_of: The declared kind of a destination, or None for a file the sources does not classify.
licence_states_an_open_source_codebase: Whether one SPDX expression is well formed and states an open-source codebase.
marker_defect: The one definition of an ill-formed block document, shared by every splice and every reader that judges one: None for a whole-file destination or a well-formed document.
nix_unsupported_shape: Why the whole Nix capability stays out of a landing, or None where the target’s crate shape supports the seed.
nix_withholding: The Nix destinations an opted-in landing withholds at this target, with the one reason, or None where the capability lands whole or nix is off.
release_driver_shape: What a release driver’s own release request is shaped like, and which job maintains it, where this release drives that driver on GitHub.
release_gate_spans: The three replaceable spans of a release workflow under the rendering that gates its request, each as its ordered begin and end marker.
render: Substitute the landing parameters into a rendered file’s bytes.
routing_block: The routing block for one checkout mode, from this binary’s embedded templates.
routing_line: The routing block’s mode line for one checkout mode.
scope_is_shaped: Whether one scope matches SCOPE_SHAPE.
splice_hooks_block: The whole .pre-commit-config.yaml content after splicing the rendered hook block.
splice_marked_block: The whole document’s bytes after splicing a marked block into it.
substitute: Every token occurrence replaced with value.
supported_pairs: Every (driver, forge) pair at which the release automation is available, in path order.