ConfigKit/docs/roadmap/sequences.md

# Sequence Identity And Cross-Tier Combination

Status: draft proposal for discussion. This document captures the requested
capabilities, settled design choices, and initial recommendations. Choices under
Open Decisions remain unsettled.
It is not an implementation plan or an implemented contract. All new syntax
shown below is illustrative.

## Motivation

Allow configuration authors to assemble collections from independently authored
records, declare their identity in configuration, and progressively extend a
baseline collection through higher layers. Authors should also be able to
replace the accumulated collection or replace individual records by identity.

The motivating example is a prompt collection whose defaults supply some
records and whose application or installation layers add or supersede records.
Across tiers, authors need three operations: whole-sequence replacement,
duplicate-rejecting append, and whole-record replacement by identity. The latter
retains unmatched lower-tier records and adds higher-tier records with new
identities. Matching records are replaced in full.

## Current Baseline

The [import contract](../config.md#explicit-imports) owns current selector,
destination, placement, and ordering behavior. An import with
`placement: sequence-item` places each selected source mapping as one item at
its effective destination. Item-local helper imports can contribute fields
before placement.

The application currently registers sequence identity through
`Definition.SequenceCompositions`, using one literal string field at each
registered mapping path. Within a strict tier, distinct identities append and
duplicate identities fail. Repeated unregistered sequences conflict. Across
precedence tiers, a higher sequence replaces the lower sequence wholesale.
The [composition contract](../config.md#strict-document-composition) owns these
rules; [layered loading](../config.md#layered-loading) owns tier precedence.

Source-backed underlays and `Definition.Defaults` have different lifecycle
roles. Underlays participate in source assembly; typed defaults are applied
later. Extending one does not automatically imply extending the other.

## Requested Capabilities

The proposed import syntax introduces an identity declaration, tentatively
named `key`, that accepts either one field or a list of fields for a compound
identity. Applications retain an optional interface for registering a collection
and imposing its identity definition. Configuration declarations must conform
to an application-owned constraint when one exists. The exact Go interface and
configuration declaration site remain to be designed.

```yaml
configkit:
  imports:
    - dir: prompts
      pattern: '*.yaml'
      destination: [prompts]
      placement: sequence-item
      key: [namespace, id]
      mode: append
```

Each selected file would contribute a record such as:

```yaml
namespace: support
id: welcome
text: Welcome
```

Here `mode: append` illustrates retaining a lower-layer collection and adding
distinct records from the higher layer. A named operation is the current
recommendation for expressing all three choices; field spelling, operation
names, and default behavior remain open. The earlier `append` or inverse
`replace` boolean proposals would need additional syntax for the third operation.

The intended results can be expressed without settling the syntax. Letters
below denote record identities; `B*` is a higher-tier record with the same
identity as `B` and a different payload. The outcomes are settled; the ordering
shown is the recommendation discussed below.

| Lower-tier sequence | Higher-tier sequence | Operation | Proposed result |
| --- | --- | --- | --- |
| `[A, B]` | `[C]` | Append | `[A, B, C]` |
| `[A, B]` | `[C]` | Whole-sequence replacement | `[C]` |
| `[A, B]` | `[B*]` | Append | Duplicate-identity error |
| `[A, B]` | `[B*]` | Whole-sequence replacement | `[B*]` |
| `[A, B]` | `[B*]` | Replacement by identity | `[A, B*]` |
| `[A, B]` | `[B*, C]` | Replacement by identity | `[A, B*, C]` |

Whole-sequence replacement discards the lower sequence. A repeated identity in the
retained combined sequence is an error under the proposed append semantics,
including when the records have identical or disjoint payloads.

## Settled Design Choices

### Whole-Record Replacement By Identity

The third cross-tier operation replaces each matching lower record with the
complete higher record, retains unmatched lower records, and adds new identities.
For example, given these lower and higher records:

```yaml
# Lower record
id: B
text: Original
temperature: 0.5
```

```yaml
# Higher record
id: B
text: Revised
```

The resulting record at this precedence boundary contains `id` and `text`.
The lower record's `temperature` is discarded. Subsequent defaulting or other
lifecycle operations retain their separately defined semantics. Recursive
field-by-field overlay of matching records is outside this selected operation.

### Optional Application-Owned Identity Constraints

Retain an interface through which applications may register a collection and
impose its unique-key definition. If the application requires `key: id`, a
configuration declaration of `key: [id, name]` is invalid: the compound key
would permit repeated `id` values with different names. The application identity
constraint applies whichever cross-tier operation configuration selects.

This is an authoritative constraint, distinct from an application default that
only fills an omission. Configuration can define identity for collections
without an application-owned constraint. The recommendation is for omitted
configuration identity to inherit a registered definition and for an explicit
matching declaration to be accepted. The final-result guarantee for
application-owned constraints is settled below; YAML-only enforcement remains
an open follow-up to OD-1.

### Final-Result Guarantee For Application Registrations

**OD-1: settled for application-owned constraints.**

> On successful loading, the final effective collections covered by application
> registrations satisfy their declared identity constraints.

Retain assembly-time identity checks and validate the final effective
collections before publication. This covers the loaded state after defaults,
inheritance, overrides, and application resolution have taken effect. Those
operations retain their separately defined combination semantics. The guarantee
does not require every intermediate callback state to satisfy the constraint
and does not extend to caller mutations after loading returns.

Keep the promise limited to the declared identity constraints. It does not by
itself require registered paths to exist or settle null and non-sequence
handling; those boundaries remain under OD-3 and OD-6. Final validation examines
effective typed values, while assembly-time checks retain their role in
validating authored contributions. A final string value alone does not prove
that the field was explicitly supplied in a source.

The existing canonical final-value representation and shared publication path
provide the implementation foundation. Diagnostics should identify the affected
collection and item paths, using available origin information without exposing
identity values. The guarantee requires neither a new mutation-history system
nor exact causal reconstruction for records supplied by defaults or callbacks.
Final enforcement for identities declared only in YAML is not settled here.

## Initial Architectural Direction

Treat identity as a property of the destination collection, even if an import
introduces its declaration. That property needs to cover both imported items
and records authored directly at the same destination. Discovering collection
declarations before composing the relevant scope would make their application
independent of which contribution happens to be assembled first.

The initial recommendation is to compose each tier's contributions strictly,
then apply one collection operation from the higher tier against the lower
sequence. An individual import would therefore express policy for its tier's
contribution at that destination. It would not sequentially erase contributions
from peer imports. Multiple agreeing declarations could coalesce; conflicting
explicit identity or operation declarations would produce errors.

For append, the ordering recommendation is to retain lower-tier item order and
then the higher tier's assembled item order. For replacement by identity, retain
the position of each replaced record and append new identities in higher-tier
order. Existing dependency-ready scheduling would continue to determine assembly
order within a tier. Whole-sequence replacement would retain only the higher
tier's assembled sequence. Ordering and the treatment of peer declarations remain
recommendations awaiting agreement, not changes to the current ordering contract.

This requires collection policy to survive import expansion through precedence
assembly. Correctness includes final values, projected item paths, source
destinations, provenance, history, sensitivity, and passive resource contexts.
Appending changes final indexes; replacing changes which contributions survive.
The implementation must account for both through the existing loading lifecycle.

## Scope And Decision Ownership

This is narrower than the deferred
[identified-record sequences proposal](future.md#1-identified-record-sequences),
which also considers recursive merging of matching records and record-specific
inheritance. Those capabilities remain separate candidates. Compound keys,
duplicate-rejecting append, and whole-record replacement by identity do not
require adopting them.

The proposal would extend identity declaration ownership and revise the
cross-tier replacement decision in
[ADR 0014](../adr/0014-sequence-placement-and-composition.md). If adopted, record
the architectural change in a superseding decision and reconcile affected
architecture policy before implementation. The current ADR and contracts remain
authoritative until that decision is made. This draft neither changes them nor
selects migration machinery.

A staged implementation plan should follow resolution of the questions below.
It should cover ordinary and layered loading, item-local scopes, source-version
processing, identity validation, precedence, and inspection together. Public API
design and test sequencing belong to that later plan.

## Open Decisions

### OD-1 Follow-Up. Final Validation For YAML-Only Identity Declarations

Application-registered collections have the settled final-result guarantee
above. Should a configuration-only identity declaration carry the same
guarantee, or govern structural assembly alone?

**Initial recommendation:** Settle this after choosing declaration scope in
OD-2 and identity-policy lifetime across replacement in OD-3. Final enforcement
needs an unambiguous effective policy for each retained collection. The
application-owned guarantee can proceed independently of this choice.

**Alternatives:** Applying the same final check to YAML-declared identities
would give both declaration forms consistent enforcement, but requires resolving
which policy survives replacement or follows nested records. Limiting YAML-only
identity to assembly keeps policy lifetime narrower, but later operations could
produce a final collection that no longer satisfies that identity declaration.

### OD-2. Where Does Configuration Declare Collection Policy?

Is `key` on a `sequence-item` import sufficient, or should configuration have a
collection declaration that also covers inline sequences, ordinary mapping
imports containing sequences, and selected overlays? How does such a declaration
address a collection inside an imported item?

**Initial recommendation:** Define one destination-scoped collection policy
model first. Evaluate a configuration-level declaration alongside import syntax
against these workflows before selecting the surface. Collection identity must
apply consistently to all participating records at that destination, including
inline records. Resolve relative destinations and item-local scopes using their
existing ownership boundaries.

**Alternative:** Limit this feature to policies declared on item imports. This
keeps the initial syntax small and directly serves imported prompt collections,
but leaves inline-only collections and overlay-authored append without a natural
declaration site. A separate collection declaration covers those cases at the
cost of more configuration structure; an import shorthand would require a clear
relationship to that declaration.

### OD-3. What Constitutes Identity, And When Can It Change?

Settle compound-key field selection, value types, missing/null components,
empty and repeated field lists, empty identity strings, and agreement between
declarations. Decide whether whole-sequence replacement can introduce a different identity
definition after discarding the lower collection when the application has imposed
no identity constraint. An application-owned constraint remains authoritative.

**Initial recommendation:** Interpret `key: id` as `[id]` and a compound key as
an ordered tuple of direct, literal string fields. Retain exact string comparison
and require every component to be present. Preserve tuple boundaries rather than
concatenating values. Identity declarations participating in append or replacement
by identity must agree; equal definitions within a scope coalesce and conflicting
definitions fail. Detailed list validation and identity changes at whole-sequence
replacement boundaries remain to be settled.

**Alternative:** Permit additional scalar types or nested field paths. These
support more schemas but add type-equality and addressing rules. A load-wide fixed
identity definition simplifies agreement across tiers; allowing a replacement to
establish a new definition gives independently authored layers more flexibility.

### OD-4. What Does An Import's Combination Choice Control?

What happens when peer imports into the same destination request different
operations, or when some contributions omit the operation? Settle ordering for
replacement by identity, including where replaced records and new identities
appear.

**Initial recommendation:** Assemble one strict sequence per destination in
each tier, then apply its agreed operation at the precedence boundary. Reject
conflicting explicit choices within a tier. This preserves peer composition and
prevents declaration or scheduling order from granting replacement authority.
Define how omitted choices inherit the effective collection operation once its
declaration site is selected. Duplicate identities among peers would remain
strict-composition errors under all three cross-tier operations. For replacement
by identity, keep matched records in their lower-tier positions and append new
identities in higher-tier order.

**Alternative:** Treat each import as a sequential collection operation.
This permits deliberate resets among imports, but a replacement can discard peer
contributions and would require an explicitly selected operation-order model.
Another alternative is to allow the choice only on a collection declaration,
which reduces conflicts between import annotations but changes the proposed
authoring surface.

For ordering, moving replaced records to their higher-tier contribution positions
is another viable choice. It gives the higher layer more influence over ordering,
but changing a record's payload could also move it relative to retained records.

### OD-5. What Is The Default Operation And Its Syntax?

Choose syntax for all three operations and decide what omission means.

**Initial recommendation:** Keep whole-sequence replacement as the cross-tier
default and make append and replacement by identity explicit. Lists can represent
complete ordered instructions as well as extensible catalogs; an explicit append
communicates the intent to retain
lower-layer records. Prefer one named operation with three values to interacting
booleans; choose names that distinguish whole-sequence from whole-record
replacement.

**Alternative:** Default identified collections to append or replacement by
identity and require explicit whole-sequence replacement. These defaults make
progressive collections concise, but retain lower-layer records unless an author
deliberately replaces the collection. Extending the boolean proposal with a
separate duplicate-handling option is possible, but introduces interacting
settings whose combinations need explanation and validation.

### OD-6. What Do Absence, Empty Sequences, And Null Mean?

Can a declaration selecting nothing affect a lower collection? How do explicit
empty sequences, null, and non-sequence values interact with all three operations?

**Initial recommendation:** An optional import selecting nothing contributes
nothing and does not clear lower values. An explicitly supplied empty sequence
with whole-sequence replacement can clear a collection; append or replacement
by identity with an empty sequence contributes no records. Preserve the
distinction between absence and supplied values. Decide
null and kind-mismatch handling explicitly rather than silently treating them
as empty sequences.

**Alternative:** Treat an explicitly declared replacement operation as a reset
even when its import selects nothing. This provides a compact clearing mechanism,
but makes an optional missing file or empty search destructive. Keeping clearing
as an explicit value avoids that coupling.

### OD-7. Which Lifecycle Stages Participate?

Do the new cross-tier operations extend only source-backed tiers, or also
`Definition.Defaults`, inheritance, CLI overrides, and application-created records?
At which boundaries
must identity be validated, including helper-supplied fields and consumed source
version headers?

**Initial recommendation:** Start with source-backed strict composition and
precedence assembly. Validate records at the relevant collection assembly
boundaries and uniqueness across each combined result. Preserve existing
typed-default and application-resolution semantics unless their inclusion is
explicitly chosen. Decide selected-overlay coverage with OD-2; it is part of
source precedence but currently has no import declaration site. Account for
item-local helpers and source-version consumption before deciding identity is
available, consistent with the current lifecycle.

The settled OD-1 guarantee covers final validation of application constraints.
Validating the final collection does not itself change how defaults, inheritance,
CLI overrides, or callbacks combine values.

**Alternative:** Extend the combination operations through defaulting,
inheritance, overrides, and application resolution. This would support
progressively extending collections across more of the lifecycle, but requires
additional combination, callback, and inspection rules. Supporting typed default
lists is a separate scope choice from extending source-backed baseline
collections; the application-owned final-result guarantee remains independent
of this combination-scope decision.