Explore features

SHOW FEATURES lets you find persisted features, inspect their formulas, and understand the dependency graph before you reuse or change anything.

Inspect a namespace

The examples use a small isolated namespace. First, create two features:

CREATE OR REPLACE FEATURES AS
PERSIST
    fm.tutorials.explore.feature1 := 1,
    fm.tutorials.explore.feature2 := 2
;

Request only the NAME column when you need a compact inventory:

SHOW FEATURES (COLUMNS (NAME)) IN FM.TUTORIALS.EXPLORE;

SHOW FEATURES always reads the current registry state. After a replacement and two additions, the same namespace immediately reflects those changes:

CREATE OR REPLACE FEATURES AS
PERSIST
    fm.tutorials.explore.feature1 := 11,
    fm.tutorials.explore.feature11 := fm.tutorials.explore.feature1 + 1,
    fm.tutorials.explore.feature12 := fm.tutorials.explore.feature1
    + fm.tutorials.explore.feature2 WITH (DESCRIPTION = 'Adds FEATURE1 and FEATURE2')
;

Filter the result

Start with the shortcut that matches your task:

The shortcuts can be combined. This query applies a name pattern inside a namespace, controls the search depth, orders the result, and limits its size:

SHOW FEATURES (COLUMNS (NAME, DATATYPE, FORMULA, STATUS)) LIKE '%EXPLORE%1' IN FM.TUTORIALS UP TO LEVEL 2 WHERE NAME NOT LIKE '%EXPLORE1%' ORDER BY NAME LIMIT 10;

Use COLUMNS when you want an exact output shape. Use INCLUDE to add derived fields to the default output, or EXCLUDE to remove fields. Because SHOW FEATURES returns tabular data, you can also use WHERE, ORDER BY, and LIMIT.

Analyze dependencies

FeatureQL stores dependencies as a directed graph. Three operators answer the common impact-analysis questions:

Create a small chain: RFEATURE0 and RFEATURE1 are inputs, and RFEATURE4 is the final derived feature.

CREATE OR REPLACE FEATURES IN FM.TUTORIALS.EXPLORE AS
PERSIST
    rfeature0 := INPUT(BIGINT),
    rfeature1 := INPUT(BIGINT),
    rfeature2 := rfeature0 + 1,
    rfeature3 := rfeature1 + rfeature2,
    rfeature4 := rfeature3 + 2
;

Find what a feature needs

IS_UPSTREAM_OF() returns the features needed to compute its argument. This query answers “What does RFEATURE3 depend on?”:

SHOW FEATURES (COLUMNS (NAME, FORMULA)) WHERE IS_UPSTREAM_OF('FM.TUTORIALS.EXPLORE.RFEATURE3')

Find what a change could affect

IS_DOWNSTREAM_OF() returns features that depend on its argument. Use it before replacing or dropping a feature:

SHOW FEATURES (COLUMNS (NAME, FORMULA)) WHERE IS_DOWNSTREAM_OF('FM.TUTORIALS.EXPLORE.RFEATURE3')

Measure graph distance

DEPTH_RELATIVE_TO() returns 0 for the reference feature, negative values for upstream dependencies, and positive values for downstream dependents:

SHOW FEATURES (
    COLUMNS (NAME, FORMULA),
    INCLUDE (DEPTH_RELATIVE_TO('FM.TUTORIALS.EXPLORE.RFEATURE3') as DEPTH_IN_DAG)
)
WHERE DEPTH_IN_DAG IS NOT NULL
ORDER BY DEPTH_IN_DAG ASC, NAME

Evaluate a complete trace

RETURN TRACE evaluates the named upstream dependency chain as well as the requested output. It returns inputs first and the requested feature last, regardless of definition order.

With local features, the ordering is easy to see:

RETURN TRACE
    f4 := f2 + f3,
    f1 := 1,
    f2 := f1 + 1,
    f3 := f1 * 2
;

With persisted features, naming only RFEATURE4 expands all five features in its upstream tree:

RETURN TRACE
    fm.tutorials.explore.rfeature4
FROM FM.TUTORIALS.EXPLORE
FOR
    fm.tutorials.explore.rfeature0 := BIND_VALUE(2),
    fm.tutorials.explore.rfeature1 := BIND_VALUE(3)
;

Use ordinary RETURN when you need only final outputs. Use RETURN TRACE when intermediate values help you understand or debug the computation. RETURN TRACE is an in-query tool — Tools in the query .

Full SHOW FEATURES shape

SHOW [ALL] [RECENT] FEATURES
  [(INCLUDE <fields> | EXCLUDE <fields> | COLUMNS <fields>)]
  [<feature_name>, ...]
  [IN <namespace> [UP TO LEVEL <n>]]
  [LIKE <pattern>]
  [WHERE <condition>]
  [ORDER BY <expression>]
  [LIMIT <n>];

Common metadata fields include:

• NAME: fully qualified feature name
• DATATYPE: FeatureQL type
• FUNCTION: top-level function used by the definition
• FORMULA: stored feature expression
• DEPENDENCIES: upstream feature names and depths
• CREATED_AT and UPDATED_AT: registry timestamps

Next

Continue with Select persisted features to evaluate the features you found, or read Registry consistency before changing a widely used definition. To inspect past definitions, see Catalog revisions . To try a change without writing HEAD, see Branch overlays . For the complete catalog selector grammar, see the SHOW syntax reference .