Indicator output tables

Indicator output tables exist only when the Job uses at least one Indicator Instance. They use a wide schema and store numeric outputs beside compact effective-state metadata.

Table names and allocation

The single-table case uses indicators. Larger configurations use indicators_1 through indicators_N.

DataCat allocates at most 1500 indicator output columns per table. One Indicator Instance is indivisible: all outputs from that instance remain together, even when unused capacity remains in the preceding table. meta.indicator_table_count records the actual allocated table count.

Base columns

Column	Type	Meaning
`open_time_ms`	`INTEGER PRIMARY KEY`	Joins the output row to `bars.open_time_ms`
`open_time_utc`	`TEXT`	Human-readable UTC mirror
`indicator_output_states_json`	`TEXT`	Effective state for every output column in the row
Generated output columns	`REAL NULL`	Numeric result or `NULL` when it cannot be computed

Use indicator_definitions.table_name and column_name to discover the generated output columns. Do not guess them from display labels.

Effective output states

final: numeric output depends only on accepted final real provider input.
provisional: output depends on at least one pending_finalization source row.
derived: output uses calculation-only normalization for at least one provider_omitted row.
warmup: output is NULL because usable history is insufficient.
awaiting_input: output is NULL because leading pending input has no previous usable close.
unavailable: output is NULL because required input is unusable.

When multiple conditions apply, precedence is unavailable > awaiting_input > warmup > provisional > derived > final.

State JSON structure

The JSON value always contains default and exceptions. Outputs not listed in an exception array use the default state.

{
  "default": "final",
  "exceptions": {
    "derived": ["i0005_sma__sma"],
    "unavailable": ["i0008_obv__obv"]
  }
}

Every output column appears in at most one exception array. Query both the numeric value and its effective state when data quality matters.

`NULL` output values

A NULL is not automatically a calculation error. It normally corresponds to warmup, awaiting_input, or unavailable. Calculation or persistence failures roll back the candidate transaction rather than committing newer bars with silently stale indicator output.