fix: preserve length modifier in varchar(n)[] array columns (#420) (#438)

Co-authored-by: vytautas.karpavicius <vytautas.karpavicius@cloudkitchens.com>

Author: vytautas-karpavicius

cmd/dump/dump_integration_test.go

@@ -151,6 +151,13 @@ func TestDumpCommand_Issue191FunctionProcedureOverload(t *testing.T) {
 	runExactMatchTest(t, "issue_191_function_procedure_overload")
 }
 
+func TestDumpCommand_Issue420VarcharArrayLengthModifier(t *testing.T) {
+	if testing.Short() {
+		t.Skip("Skipping integration test in short mode")
+	}
+	runExactMatchTest(t, "issue_420_varchar_array_length_modifier")
+}
+
 // Reproduces a bug where a column declared as `name` is dumped as `char[]`.
 // The inspector classifies any base type with pg_type.typelem <> 0 as an array,
 // but the `name` type has typelem = 18 (the OID of "char") despite not being an array.

ir/normalize.go

@@ -1135,9 +1135,15 @@ func normalizePostgreSQLType(input string) string {
 	// Handle direct type names
 	typeName := input
 
+	// Decompose into base name and optional modifier+suffix (e.g., "bpchar(10)[]" -> "bpchar" + "(10)[]").
+	// When there's no modifier, rest is empty and this is equivalent to a plain exact-match lookup.
+	baseName, rest := typeName, ""
+	if idx := strings.Index(typeName, "("); idx != -1 {
+		baseName, rest = typeName[:idx], typeName[idx:]
+	}
 	// Check if we have a direct mapping
-	if normalized, exists := postgresTypeNormalization[typeName]; exists {
-		return normalized
+	if normalized, exists := postgresTypeNormalization[baseName]; exists {
+		return normalized + rest
 	}
 
 	// Remove pg_catalog prefix for unmapped types
@@ -1480,4 +1486,3 @@ func findArrayClose(expr string, startIdx int) int {
 
 	return -1 // Not found
 }
-

ir/queries/queries.sql

@@ -85,11 +85,12 @@ WITH column_base AS (
                 -- Array types: apply same schema qualification logic to element type
                 -- Use typcategory = 'A' rather than typelem <> 0; the latter is true
                 -- for non-array fixed-length types like name (typelem points to char).
+                -- Use format_type to preserve typmod for element types (e.g., varchar(128)[] for character varying(128)[])
                 CASE
-                    WHEN en.nspname = 'pg_catalog' THEN et.typname || '[]'
-                    WHEN en.nspname = c.table_schema THEN et.typname || '[]'
-                    ELSE en.nspname || '.' || et.typname || '[]'
-                END
+                    WHEN en.nspname = 'pg_catalog' THEN et.typname
+                    WHEN en.nspname = c.table_schema THEN et.typname
+                    ELSE en.nspname || '.' || et.typname
+                END || COALESCE(substring(format_type(a.atttypid, a.atttypmod) FROM '\([^)]*\)'), '') || '[]'
             WHEN dt.typtype = 'b' THEN
                 -- Non-array base types: qualify if not in pg_catalog or table's schema
                 -- Use format_type to preserve typmod for extension types (e.g., vector(384) for pgvector)
@@ -203,11 +204,12 @@ WITH column_base AS (
                 -- Array types: apply same schema qualification logic to element type
                 -- Use typcategory = 'A' rather than typelem <> 0; the latter is true
                 -- for non-array fixed-length types like name (typelem points to char).
+                -- Use format_type to preserve typmod for element types (e.g., varchar(128)[] for character varying(128)[])
                 CASE
-                    WHEN en.nspname = 'pg_catalog' THEN et.typname || '[]'
-                    WHEN en.nspname = c.table_schema THEN et.typname || '[]'
-                    ELSE en.nspname || '.' || et.typname || '[]'
-                END
+                    WHEN en.nspname = 'pg_catalog' THEN et.typname
+                    WHEN en.nspname = c.table_schema THEN et.typname
+                    ELSE en.nspname || '.' || et.typname
+                END || COALESCE(substring(format_type(a.atttypid, a.atttypmod) FROM '\([^)]*\)'), '') || '[]'
             WHEN dt.typtype = 'b' THEN
                 -- Non-array base types: qualify if not in pg_catalog or table's schema
                 -- Use format_type to preserve typmod for extension types (e.g., vector(384) for pgvector)

ir/queries/queries.sql.go

@@ -0,0 +1,10 @@
+{
+  "name": "issue_420_varchar_array_length_modifier",
+  "description": "pgschema dump silently drops the length modifier from varchar(n)[] array columns, emitting varchar[] instead of varchar(128)[]",
+  "source": "https://github.com/pgplex/pgschema/issues/420",
+  "notes": [
+    "The SQL query for array types used et.typname || '[]' which discards the atttypmod",
+    "The fix uses format_type(a.atttypid, a.atttypmod) to extract the modifier (e.g., (128))",
+    "Also covers character(n)[] which had the same bug (bpchar typname)"
+  ]
+}

testdata/dump/issue_420_varchar_array_length_modifier/manifest.json

@@ -0,0 +1,33 @@
+--
+-- PostgreSQL database dump
+--
+
+SET statement_timeout = 0;
+SET lock_timeout = 0;
+SET client_encoding = 'UTF8';
+SET standard_conforming_strings = on;
+SET check_function_bodies = false;
+SET client_min_messages = warning;
+SET row_security = off;
+
+--
+-- Name: items; Type: TABLE; Schema: public; Owner: -
+--
+
+CREATE TABLE public.items (
+    id integer NOT NULL,
+    name character varying(64),
+    tags character varying(128)[],
+    codes character(10)[]
+);
+
+--
+-- Name: items items_pkey; Type: CONSTRAINT; Schema: public; Owner: -
+--
+
+ALTER TABLE ONLY public.items
+    ADD CONSTRAINT items_pkey PRIMARY KEY (id);
+
+--
+-- PostgreSQL database dump complete
+--

testdata/dump/issue_420_varchar_array_length_modifier/pgdump.sql

@@ -0,0 +1,20 @@
+--
+-- pgschema database dump
+--
+
+-- Dumped from database version PostgreSQL 18.0
+-- Dumped by pgschema version 1.9.0
+
+
+--
+-- Name: items; Type: TABLE; Schema: -; Owner: -
+--
+
+CREATE TABLE IF NOT EXISTS items (
+    id integer,
+    name varchar(64),
+    tags varchar(128)[],
+    codes character(10)[],
+    CONSTRAINT items_pkey PRIMARY KEY (id)
+);
+

testdata/dump/issue_420_varchar_array_length_modifier/pgschema.sql

@@ -0,0 +1,13 @@
+--
+-- Test case for GitHub issue #420: varchar(n)[] length modifier silently dropped in dump
+--
+-- The length modifier is lost when dumping array columns of character types
+-- with a length constraint: varchar(128)[] is emitted as varchar[].
+--
+
+CREATE TABLE items (
+    id      integer PRIMARY KEY,
+    name    varchar(64),
+    tags    varchar(128)[],
+    codes   character(10)[]
+);