feat(source-stripe): expose supports_realtime_sync in catalog metadata

[kdhillon-stripe]

Summary

The Stripe source discover() had two independent discovery paths that each walked the OpenAPI spec separately, applied different inclusion logic, produced different output shapes, and were then joined by a fragile name-based intersection:

buildResourceRegistry(spec)            SpecParser.parse(spec)
  discoverListEndpoints()                discoverAllowedTables()
  - finds ~90 top-level listable           discoverListableResourceIds(includeNested:true)
  - filters by EXCLUDED_TABLES               - finds ~112 listable (incl. nested)
  - builds listFn/retrieveFn              discoverWebhookUpdatableResourceIds()
  - outputs: Record<name, ResourceConfig>    - finds ~63 with webhooks
                                             - intersects: ~59 resource IDs
                                             - resolves table names via aliases
                                             - outputs: ParsedResourceTable[]
         \                                /
          \                              /
           catalogFromOpenApi(tables, registry)
             iterates registry entries
             looks up table by name from a separate Map
             result: catalog streams

Neither side owned both concerns: buildResourceRegistry was the source of truth for what can be synced (it has the runtime list/retrieve functions), while SpecParser was the source of truth for what should be synced (webhook filter + schema). The two were glued together by a name-based Map.get() that could silently produce streams without schemas ("ghost tables") or quietly drop tables when the two sets disagreed.

Solution

Establish a single linear pipeline where each step feeds the next, rather than two parallel paths joined at the end.

Before

flowchart TB
    spec[OpenAPI spec]

    spec --> left["buildResourceRegistry(spec)"]
    spec --> right["SpecParser.parse(spec)"]

    left --> reg["registry — runtime fns per table"]
    right --> tables["parsed.tables — schemas per table"]

    reg --> cfo["catalogFromOpenApi(tables, registry)"]
    tables --> cfo
    cfo -->|"join by name"| catalog["catalog.streams"]

Loading

Two independent walks of the spec, joined by name at the end.

After

flowchart TB
    spec[OpenAPI spec]

    spec --> ps["parser.parseSyncable(spec, {excluded})"]
    ps --> parsed["parsed.tables — syncable tables with schemas"]

    parsed --> brr["buildResourceRegistry(spec, ..., parsedTables)"]
    spec --> brr
    brr --> reg["registry — runtime fns + schema per entry"]

    reg --> cfo["catalogFromOpenApi(registry)"]
    cfo --> catalog["catalog.streams"]

Loading

Single linear flow. No parallel discovery, no name-based join.

What changed

1. SpecParser.parseSyncable() — new method that fuses discoverSyncableTables + parse into one call. The canonical "what should be synced" decision happens once, here.

2. parsedTable on ResourceConfig — each registry entry now carries its own parsed schema, attached at build time. The schema travels with the runtime functions instead of being looked up by name later.

3. catalogFromOpenApi(registry) — takes one argument instead of two. No more Map.get() join. Each entry already has everything needed to emit a stream. Throws if parsedTable is missing (ghost-table guard moved here).

[Copilot]

Pull request overview

Adds per-stream catalog metadata to indicate whether a Stripe table is updated via real-time webhooks, enabling destinations to distinguish webhook-backed streams from backfill-only streams.

Changes:

• Introduces REALTIME_SYNC_TABLES as a canonical set of webhook-updated table names.
• Sets stream.metadata.supports_realtime_sync in catalogFromOpenApi() based on REALTIME_SYNC_TABLES membership.
• Adds unit tests covering one real-time stream (subscriptions) and one list-only stream (reporting_report_types).

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
packages/source-stripe/src/types.ts	Adds REALTIME_SYNC_TABLES to enumerate tables receiving webhook-driven updates.
packages/source-stripe/src/catalog.ts	Populates metadata.supports_realtime_sync for each discover stream.
packages/source-stripe/src/catalog.test.ts	Adds tests asserting the flag is true for a webhook-backed stream and false for a list-only stream.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.