Skip to content

mcp: Cover mz_mcp_data_products view semantics #37636

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
bobbyiliev merged 1 commit into from
Aug 3, 2026
Merged

mcp: Cover mz_mcp_data_products view semantics #37636

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
226 changes: 226 additions & 0 deletions test/sqllogictest/rbac_mcp_agent.slt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1003,6 +1003,232 @@ ALTER ROLE agent RESET restrict_to_user_objects;
----
COMPLETE 0

# =============================================================================
# Data product view semantics (mz_mcp_data_products / _details)
#
# The RBAC cases above exercise privilege filtering. These cases pin the
# view logic itself: which objects qualify, how the description is chosen,
# and how column types map into the JSON Schema. They run as a
# fully-privileged owner so the object set is deterministic and independent
# of the agent-role grants above.
# =============================================================================

statement ok
CREATE SCHEMA dp_semantics;

statement ok
SET search_path = dp_semantics;

# A materialized view qualifies as a data product with no index required.
statement ok
CREATE MATERIALIZED VIEW mv_plain AS SELECT 1 AS id;

# A plain (non-indexed) view does NOT qualify: reading it would trigger a
# full recompute.
statement ok
CREATE VIEW v_unindexed AS SELECT 1 AS id;

# A view qualifies only once it has an index.
statement ok
CREATE VIEW v_indexed AS SELECT 1 AS id, 'a' AS label;

statement ok
CREATE INDEX v_indexed_idx ON v_indexed (id);

query T
SELECT object_name FROM mz_internal.mz_mcp_data_products
WHERE object_name LIKE '%"dp_semantics"%' ORDER BY object_name;
----
"materialize"."dp_semantics"."mv_plain"
"materialize"."dp_semantics"."v_indexed"

# Multiple indexes on one object must not duplicate its row (SELECT DISTINCT).
statement ok
CREATE INDEX v_indexed_idx2 ON v_indexed (label);

query T
SELECT object_name FROM mz_internal.mz_mcp_data_products
WHERE object_name = '"materialize"."dp_semantics"."v_indexed"';
----
"materialize"."dp_semantics"."v_indexed"

# -----------------------------------------------------------------------------
# Description precedence: index comment wins over object comment (COALESCE).
# -----------------------------------------------------------------------------

# No comments: description is NULL.
query T
SELECT description FROM mz_internal.mz_mcp_data_products
WHERE object_name = '"materialize"."dp_semantics"."mv_plain"';
----
NULL

# Object comment only: description is the object comment.
statement ok
COMMENT ON MATERIALIZED VIEW mv_plain IS 'the object comment';

query T
SELECT description FROM mz_internal.mz_mcp_data_products
WHERE object_name = '"materialize"."dp_semantics"."mv_plain"';
----
the object comment

# Index comment present: it takes precedence over the object comment. Uses a
# dedicated single-index view so the COALESCE precedence is unambiguous (with
# multiple indexes the DISTINCT can surface one row per differing description).
statement ok
CREATE VIEW v_desc AS SELECT 1 AS id;

statement ok
CREATE INDEX v_desc_idx ON v_desc (id);

statement ok
COMMENT ON VIEW v_desc IS 'object comment on view';

statement ok
COMMENT ON INDEX v_desc_idx IS 'index comment wins';

query T
SELECT description FROM mz_internal.mz_mcp_data_products
WHERE object_name = '"materialize"."dp_semantics"."v_desc"';
----
index comment wins

# -----------------------------------------------------------------------------
# JSON Schema type mapping in mz_mcp_data_product_details.
# -----------------------------------------------------------------------------

statement ok
CREATE MATERIALIZED VIEW mv_types AS
SELECT
1::int AS c_int,
1.5::double AS c_double,
true AS c_bool,
'x'::text AS c_text,
'2024-01-01'::date AS c_date,
'{}'::jsonb AS c_jsonb,
'\x00'::bytea AS c_bytea,
'12:34:56'::time AS c_time,
'2024-01-01 12:00:00'::timestamp AS c_ts,
'2024-01-01 12:00:00+00'::timestamptz AS c_tstz,
'00000000-0000-0000-0000-000000000000'::uuid AS c_uuid;

# Numeric columns map to JSON Schema type "number".
query T
SELECT schema->'properties'->'c_int'->>'type' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
number

query T
SELECT schema->'properties'->'c_double'->>'type' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
number

# Boolean maps to "boolean".
query T
SELECT schema->'properties'->'c_bool'->>'type' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
boolean

# text falls through to the "string" default.
query T
SELECT schema->'properties'->'c_text'->>'type' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string

# date is a string with a format hint.
query TT
SELECT schema->'properties'->'c_date'->>'type', schema->'properties'->'c_date'->>'format'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string date

# jsonb maps to a nested "object".
query T
SELECT schema->'properties'->'c_jsonb'->>'type' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
object

# bytea is the most distinctive branch: string with the two extra content
# hints so a JSON Schema consumer knows to Base64-decode the value.
query TTT
SELECT
schema->'properties'->'c_bytea'->>'type',
schema->'properties'->'c_bytea'->>'contentEncoding',
schema->'properties'->'c_bytea'->>'contentMediaType'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string base64 application/octet-stream

# time is a string with format=time.
query TT
SELECT schema->'properties'->'c_time'->>'type', schema->'properties'->'c_time'->>'format'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string time

# timestamp and timestamptz both match the `ilike 'timestamp%'` branch and
# map to string with format=date-time.
query TT
SELECT schema->'properties'->'c_ts'->>'type', schema->'properties'->'c_ts'->>'format'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string date-time

query TT
SELECT schema->'properties'->'c_tstz'->>'type', schema->'properties'->'c_tstz'->>'format'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string date-time

# uuid is a string with format=uuid.
query TT
SELECT schema->'properties'->'c_uuid'->>'type', schema->'properties'->'c_uuid'->>'format'
FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
string uuid

# -----------------------------------------------------------------------------
# Column comments surface as the property description; absent otherwise.
# -----------------------------------------------------------------------------

statement ok
COMMENT ON COLUMN mv_types.c_int IS 'the primary id';

query T
SELECT schema->'properties'->'c_int'->>'description' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
the primary id

# A column with no comment has no description key (jsonb_strip_nulls removes it).
query T
SELECT schema->'properties'->'c_double' ? 'description' FROM mz_internal.mz_mcp_data_product_details
WHERE object_name = '"materialize"."dp_semantics"."mv_types"';
----
false

# -----------------------------------------------------------------------------
# Cleanup for this section.
# -----------------------------------------------------------------------------

statement ok
SET search_path = public;

statement ok
DROP SCHEMA dp_semantics CASCADE;

# =============================================================================
# Cleanup
# =============================================================================
Expand Down
Loading