Merge pull request #3556 from mathesar-foundation/permissions_stub

Add `columns.list` function

config/settings/common_settings.py

@@ -65,7 +65,8 @@ def pipe_delim(pipe_string):
 ROOT_URLCONF = "config.urls"
 
 MODERNRPC_METHODS_MODULES = [
- 'mathesar.rpc.connections'
+ 'mathesar.rpc.connections',
+ 'mathesar.rpc.columns'
 ]
 
 TEMPLATES = [

db/columns/operations/select.py

@@ -3,11 +3,48 @@
 from sqlalchemy import and_, asc, cast, select, text, exists, Identity
 
 from db.columns.exceptions import DynamicDefaultWarning
-from db.connection import execute_msar_func_with_engine
+from db.connection import execute_msar_func_with_engine, exec_msar_func
 from db.tables.operations.select import reflect_table_from_oid
 from db.utils import execute_statement, get_pg_catalog_table
 
 
+def get_column_info_for_table(table, conn):
+ """
+ Return a list of dictionaries describing the columns of the table.
+
+ The `table` can be given as either a "qualified name", or an OID.
+ The OID is the preferred identifier, since it's much more robust.
+
+ The returned list contains dictionaries of the following form:
+
+ {
+ "id": <int>,
+ "name": <str>,
+ "type": <str>,
+ "type_options": {
+ "precision": <int>,
+ "scale": <int>,
+ "fields": <str>,
+ "length": <int>,
+ "item_type": <str>,
+ },
+ "nullable": <bool>,
+ "primary_key": <bool>,
+ "valid_target_types": [<str>, <str>, ..., <str>]
+ "default": {"value": <str>, "is_dynamic": <bool>},
+ "has_dependents": <bool>,
+ "description": <str>
+ }
+
+ The fields of the "type_options" dictionary are all optional,
+ depending on the "type" value.
+
+ Args:
+ table: The table for which we want column info.
+ """
+ return exec_msar_func(conn, 'get_column_info', table).fetchone()[0]
+
+
 def get_column_description(oid, attnum, engine):
  cursor = execute_msar_func_with_engine(engine, 'col_description', oid, attnum)
  row = cursor.fetchone()

db/connection.py

@@ -40,6 +40,21 @@ def execute_msar_func_with_psycopg2_conn(conn, func_name, *args):
  return conn.execute(stmt)
 
 
+def exec_msar_func(conn, func_name, *args):
+ """
+ Execute an msar function using a psycopg (3) connection.
+
+ Args:
+ conn: a psycopg connection
+ func_name: The unqualified msar_function name (danger; not sanitized)
+ *args: The list of parameters to pass
+ """
+ # Returns a cursor
+ return conn.execute(
+ f"SELECT msar.{func_name}({','.join(['%s']*len(args))})", args
+ )
+
+
 def load_file_with_engine(engine, file_handle):
  """Run an SQL script from a file, using psycopg."""
  conn_str = str(engine.url)

db/sql/0_msar.sql

@@ -131,6 +131,23 @@ Wraps the `?` jsonb operator for improved readability.
 $$ LANGUAGE plpgsql;
 
 
+CREATE OR REPLACE FUNCTION msar.schema_exists(schema_name text) RETURNS boolean AS $$/*
+Return true if the given schema exists in the current database, false otherwise.
+*/
+SELECT EXISTS (SELECT 1 FROM pg_namespace WHERE nspname=schema_name);
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
+CREATE OR REPLACE FUNCTION __msar.get_schema_oid(sch_name text) RETURNS oid AS $$/*
+Return the OID of a schema, if it can be diretly found from a name.
+
+Args :
+ sch_name: The name of the schema.
+*/
+SELECT CASE WHEN msar.schema_exists(sch_name) THEN sch_name::regnamespace::oid END;
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
 CREATE OR REPLACE FUNCTION __msar.get_schema_name(sch_id oid) RETURNS TEXT AS $$/*
 Return the name for a given schema, quoted as appropriate.
 
@@ -468,17 +485,180 @@ AND attrelid = msar.get_relation_oid(sch_name, rel_name);
 $$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
 
 
-CREATE OR REPLACE FUNCTION msar.column_exists(tab_id oid, col_name text) RETURNS boolean AS $$/*
-Return true if the given column exists in the table, false otherwise.
+CREATE OR REPLACE FUNCTION
+msar.get_interval_fields(typ_mod integer) RETURNS text AS $$/*
+Return the string giving the fields for an interval typmod integer.
+
+This logic is ported from the relevant PostgreSQL source code, reimplemented in SQL. See the
+`intervaltypmodout` function at
+https://doxygen.postgresql.org/backend_2utils_2adt_2timestamp_8c.html
+
+Args:
+ typ_mod: The atttypmod from the pg_attribute table. Should be valid for the interval type.
 */
-SELECT EXISTS (SELECT 1 FROM pg_attribute WHERE attrelid=tab_id AND attname=col_name);
+SELECT CASE (typ_mod >> 16 & 32767)
+ WHEN 1 << 2 THEN 'year'
+ WHEN 1 << 1 THEN 'month'
+ WHEN 1 << 3 THEN 'day'
+ WHEN 1 << 10 THEN 'hour'
+ WHEN 1 << 11 THEN 'minute'
+ WHEN 1 << 12 THEN 'second'
+ WHEN (1 << 2) | (1 << 1) THEN 'year to month'
+ WHEN (1 << 3) | (1 << 10) THEN 'day to hour'
+ WHEN (1 << 3) | (1 << 10) | (1 << 11) THEN 'day to minute'
+ WHEN (1 << 3) | (1 << 10) | (1 << 11) | (1 << 12) THEN 'day to second'
+ WHEN (1 << 10) | (1 << 11) THEN 'hour to minute'
+ WHEN (1 << 10) | (1 << 11) | (1 << 12) THEN 'hour to second'
+ WHEN (1 << 11) | (1 << 12) THEN 'minute to second'
+END;
 $$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
 
 
-CREATE OR REPLACE FUNCTION msar.schema_exists(schema_name text) RETURNS boolean AS $$/*
-Return true if the given schema exists in the current database, false otherwise.
+CREATE OR REPLACE FUNCTION
+msar.get_type_options(typ_id regtype, typ_mod integer, typ_ndims integer) RETURNS jsonb AS $$/*
+Return the type options calculated from a type, typmod pair.
+
+This function uses a number of hard-coded constants. The form of the returned object is determined
+by the input type, but the keys will be a subset of:
+ precision: the precision of a numeric or interval type. See PostgreSQL docs for details.
+ scale: the scale of a numeric type
+ fields: See PostgreSQL documentation of the `interval` type.
+ length: Applies to "text" types where the user can specify the length.
+ item_type: Gives the type of array members for array-types
+
+Args:
+ typ_id: an OID or valid type representing string will work here.
+ typ_mod: The integer corresponding to the type options; see pg_attribute catalog table.
+ typ_ndims: Used to determine whether the type is actually an array without an extra join.
 */
-SELECT EXISTS (SELECT 1 FROM pg_namespace WHERE nspname=schema_name);
+SELECT nullif(
+ CASE
+ WHEN typ_id = ANY('{numeric, _numeric}'::regtype[]) THEN
+ jsonb_build_object(
+ -- This calculation is modified from the relevant PostgreSQL source code. See the function
+ -- numeric_typmod_precision(int32) at
+ -- https://doxygen.postgresql.org/backend_2utils_2adt_2numeric_8c.html
+ 'precision', ((nullif(typ_mod, -1) - 4) >> 16) & 65535,
+ -- This calculation is from numeric_typmod_scale(int32) at the same location
+ 'scale', (((nullif(typ_mod, -1) - 4) & 2047) # 1024) - 1024
+ )
+ WHEN typ_id = ANY('{interval, _interval}'::regtype[]) THEN
+ jsonb_build_object(
+ 'precision', nullif(typ_mod & 65535, 65535),
+ 'fields', msar.get_interval_fields(typ_mod)
+ )
+ WHEN typ_id = ANY('{bpchar, _bpchar, varchar, _varchar}'::regtype[]) THEN
+ -- For char and varchar types, the typemod is equal to 4 more than the set length.
+ jsonb_build_object('length', nullif(typ_mod, -1) - 4)
+ WHEN typ_id = ANY(
+ '{bit, varbit, time, timetz, timestamp, timestamptz}'::regtype[]
+ || '{_bit, _varbit, _time, _timetz, _timestamp, _timestamptz}'::regtype[]
+ ) THEN
+ -- For all these types, the typmod is equal to the precision.
+ jsonb_build_object(
+ 'precision', nullif(typ_mod, -1)
+ )
+ ELSE jsonb_build_object()
+ END
+ || CASE
+ WHEN typ_ndims>0 THEN
+ -- This string wrangling is debatably dubious, but avoids a slow join.
+ jsonb_build_object('item_type', rtrim(typ_id::regtype::text, '[]'))
+ ELSE '{}'
+ END,
+ '{}'
+)
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
+CREATE OR REPLACE FUNCTION msar.get_valid_target_type_strings(typ_id regtype) RETURNS jsonb AS $$/*
+Given a source type, return the target types for which Mathesar provides a casting function.
+
+Args:
+ typ_id: The type we're casting from.
+*/
+
+SELECT jsonb_agg(prorettype::regtype::text)
+FROM pg_proc
+WHERE
+ pronamespace=__msar.get_schema_oid('mathesar_types')
+ AND proargtypes[0]=typ_id
+ AND left(proname, 5) = 'cast_';
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
+CREATE OR REPLACE FUNCTION msar.has_dependents(rel_id oid, att_id smallint) RETURNS boolean AS $$/*
+Return a boolean according to whether the column identified by the given oid, attnum pair is
+referenced (i.e., would dropping that column require CASCADE?).
+
+Args:
+ rel_id: The relation of the attribute.
+ att_id: The attnum of the attribute in the relation.
+*/
+SELECT EXISTS (
+ SELECT 1 FROM pg_depend WHERE refobjid=rel_id AND refobjsubid=att_id AND deptype='n'
+);
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
+CREATE OR REPLACE FUNCTION msar.get_column_info(tab_id regclass) RETURNS jsonb AS $$/*
+Given a table identifier, return an array of objects describing the columns of the table.
+
+Each returned JSON object in the array will have the form:
+ {
+ "id": <int>,
+ "name": <str>,
+ "type": <str>,
+ "type_options": <obj>,
+ "nullable": <bool>,
+ "primary_key": <bool>,
+ "default": {"value": <str>, "is_dynamic": <bool>},
+ "has_dependents": <bool>,
+ "description": <str>
+ }
+
+The `type_options` object is described in the docstring of `msar.get_type_options`. The `default`
+object has the keys:
+ value: A string giving the value (as an SQL expression) of the default.
+ is_dynamic: A boolean giving whether the default is (likely to be) dynamic.
+*/
+SELECT jsonb_agg(
+ jsonb_build_object(
+ 'id', attnum,
+ 'name', attname,
+ 'type', CASE WHEN attndims>0 THEN '_array' ELSE atttypid::regtype::text END,
+ 'type_options', msar.get_type_options(atttypid, atttypmod, attndims),
+ 'nullable', NOT attnotnull,
+ 'primary_key', COALESCE(pgi.indisprimary, false),
+ 'default',
+ nullif(
+ jsonb_strip_nulls(
+ jsonb_build_object(
+ 'value',
+ CASE
+ WHEN attidentity='' THEN pg_get_expr(adbin, tab_id)
+ ELSE 'identity'
+ END,
+ 'is_dynamic', msar.is_default_possibly_dynamic(tab_id, attnum)
+ )
+ ),
+ jsonb_build_object()
+ ),
+ 'has_dependents', msar.has_dependents(tab_id, attnum),
+ 'description', msar.col_description(tab_id, attnum)
+ )
+)
+FROM pg_attribute pga
+ LEFT JOIN pg_index pgi ON pga.attrelid=pgi.indrelid AND pga.attnum=ANY(pgi.indkey)
+ LEFT JOIN pg_attrdef pgd ON pga.attrelid=pgd.adrelid AND pga.attnum=pgd.adnum
+WHERE pga.attrelid=tab_id AND pga.attnum > 0 and NOT attisdropped;
+$$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;
+
+
+CREATE OR REPLACE FUNCTION msar.column_exists(tab_id oid, col_name text) RETURNS boolean AS $$/*
+Return true if the given column exists in the table, false otherwise.
+*/
+SELECT EXISTS (SELECT 1 FROM pg_attribute WHERE attrelid=tab_id AND attname=col_name);
 $$ LANGUAGE SQL RETURNS NULL ON NULL INPUT;