Support compact tuple-like arrays for array-of-model types

[naferok]

Problem
Data sources (tags, metrics, key-value pairs, etc.) use very compact array-of-arrays format for efficiency:
[ ["tag-w-100", 100], ["tag-w-10", 0], ["priority-high", 85] ]

We want to map this directly to a Taxi model like:
`model TagWeight {
name : String
weight : Int
}

model Content inherits TagWeight[]`

Currently, when Orbital ingests such JSON, it falls back to lang.taxi.Any and logs a warning
WARN com.orbitalhq.models.TypedInstance$Companion - [ ] - Provided value is a collection, but the declared type lang.taxi.Any is not

Benefits
Much smaller payloads (important for high-cardinality tags / metrics)

Current workarounds (not ideal)
Force source to send full object array → often impossible (legacy APIs, size constraints, etc.)

[martypitt]

I think this is a good idea - but worth clarifying, as the model declaration doesn't match the way this would be modelled.

I think in Taxi, this would end up looking something like:

type TagName inherits String
type TagWeight inherits Int

model TagWeightPair inherits [TagName,TagWeight]  // ie., a single tuple like ["tag-w-100", 100]

// Then, if returned from an operation
service MyApi {
   operation foo():TagWeightPair[] // eg [ ["tag-w-100", 100], ["tag-w-10", 0], ["priority-high", 85] ]
}

There are challenges here within the current implementation of Taxi's type system - specifically, we prevent inheritence of arrays, so you can't declare:

model Foo inherits Bar[]

I'd need to think about the syntax for this, but am generally supportive, as it's a common use-case.

Just to clarify - do you have an actual use-case for this?

[martypitt]

Playground sketch is here

[naferok]

Here’s a list of use cases where native support for compact array-of-arrays / tuples would bring real, measurable value.

For me personally, the highest-priority case right now is ClickHouse integration via HTTP JSONCompact — because this is the integration bus I actively use for analytics, event tracking, monitoring dashboards, and user behavior streams.

Reducing response payload size by 40–70% (especially with thousands or millions of rows) directly helps deliver faster APIs, lower hosting/traffic costs, better performance for clients (especially mobile/edge), and less memory pressure during processing.

My Personal: ClickHouse HTTP JSONCompact Queries

Use case: Exposing analytical queries via HTTP API — telemetry, logs, events, monitoring dashboards.

type Filter inherits String

type EventTime   inherits Instant
type UserId      inherits String
type EventType   inherits String
type Value       inherits Decimal

model EventRow inherits [EventTime, UserId, EventType, Value]

service AnalyticsApi {
   // Returns only the "data" part as array-of-arrays
   operation Events(): EventRow[]
}

events
[
  ["2026-01-20T14:35:22Z", "usr_abc123", "purchase",  89.99],
  ["2026-01-20T14:36:45Z", "usr_abc123", "view_item", 0.00],
  ["2026-01-21T09:12:10Z", "usr_abc123", "add_to_cart", 0.00],
  ["2026-01-21T09:15:33Z", "usr_abc123", "purchase", 149.50]
]

Ad / Recommendation Feature Vectors

type UserId inherits String
type FeatureName inherits String
type Value inherits Decimal
model FeaturePair inherits [FeatureName , Value ]    

service RecommendationApi {
   operation getUserFeatures(UserId): FeaturePair[]
}

user features
[
  ["age_group", 4.0],
  ["location_cluster", 17],
  ["last_category_id", 42.0],
  ["recency_days", 3.2],
  ["avg_order_value", 67.45]
]

Polygon / MultiPolygon Boundaries (GIS / Mapping APIs)

Use case: Delivery zones, country borders, heat-map regions.

type City inherits String

type Latitude  inherits Decimal
type Longitude inherits Decimal
type IsDliverable Inherits Boolean

model Point       inherits [Longitude, Latitude ]      // [lon, lat]
model Ring        inherits Point[]                 // closed ring
model Polygon     inherits Ring[]                  // exterior + holes
model DeliveryZone [IsDliverable,Polygon]

service GeometryApi {
   operation getDeliveryZone(City): DeliveryZone
}

delivery zone
[
  true,
  [
    [ // ring
      [2.294, 48.858],
      [2.300, 48.865],
      [2.310, 48.870],
      [2.320, 48.865],
      [2.315, 48.855],
      [2.294, 48.858]  
    ],
    [ // hole 
      [2.305, 48.862],
      [2.308, 48.865],
      [2.310, 48.862],
      [2.305, 48.862]
    ]
  ]
]

Sensor / IoT Telemetry Streams

Use case: IoT devices, industrial sensors, autonomous vehicles

type DeviceId inherits String

type Timestamp inherits Long
type Temperature inherits Decimal
type Humidity  inherits Decimal
type Pressure inherits Decimal

model ReadingTuple inherits [Timestamp , Temperature , Humidity  , Pressure]  

service DeviceTelemetry {
   operation getLatestReadings(DeviceId): ReadingTuple[]
   operation streamReadings(DeviceId): Stream<ReadingTuple>
}

some with null value
[
  [1737574800000, 23.4,  45.2, null],
  [1737574900000, 23.5,  44.8, 0.0],
  [1737575000000, 23.6,  46.1, 0.0],
  [1737575100000, 23.7,  47.0, 0.0]
]

[martypitt]

Yep, those are some pretty great examples 😁