feat(cli): add --dry-run to convert, merge and split

Read and validate inputs, apply filtering/merging in memory, validate the
output target (format, --package, --schema, --extra-fields) and print the
plan to stdout without writing any output. Closes #35.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

cmd/relspec/convert.go

@@ -3,6 +3,7 @@ package main
 import (
 	stdjson "encoding/json"
 	"fmt"
+	"io"
 	"os"
 	"strings"
 	"time"
@@ -57,6 +58,7 @@ var (
 	convertNullableArrays  string
 	convertContinueOnError bool
 	convertExtraFields     string
+	convertDryRun          bool
 )
 
 var convertCmd = &cobra.Command{
@@ -185,6 +187,8 @@ func init() {
 	convertCmd.Flags().BoolVar(&convertContinueOnError, "continue-on-error", false, "Prepend \\set ON_ERROR_STOP off to generated SQL so psql continues past errors (pgsql output only)")
 	convertCmd.Flags().StringVar(&convertExtraFields, "extra-fields", "", "Path to JSON file containing extra Bun model fields to inject (bun output only); fields support target_table, name, type, bun_tag, json_tag, comment")
 
+	convertCmd.Flags().BoolVar(&convertDryRun, "dry-run", false, "Read and validate the input and print the plan without writing any output")
+
 	err := convertCmd.MarkFlagRequired("from")
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "Error marking from flag as required: %v\n", err)
@@ -240,6 +244,22 @@ func runConvert(cmd *cobra.Command, args []string) error {
 	}
 	fmt.Fprintf(os.Stderr, "      Found: %d table(s)\n\n", totalTables)
 
+	if convertDryRun {
+		if err := validateWriteTarget(db, convertTargetType, convertPackageName, convertSchemaFilter, convertExtraFields); err != nil {
+			return fmt.Errorf("dry run validation failed: %w", err)
+		}
+		out := outWriter(cmd)
+		fmt.Fprintf(out, "RelSpec convert plan (dry run - nothing written):\n")
+		fmt.Fprintf(out, "  Input:  %s database '%s'\n", convertSourceType, db.Name)
+		fmt.Fprintf(out, "  Output: %s -> %s\n", convertTargetType, convertTargetPath)
+		if convertSchemaFilter != "" {
+			fmt.Fprintf(out, "  Schema filter: %s\n", convertSchemaFilter)
+		}
+		printDryRunPlan(out, db)
+		fmt.Fprintf(os.Stderr, "=== Dry run complete: no output written ===\n\n")
+		return nil
+	}
+
 	// Write to target format
 	fmt.Fprintf(os.Stderr, "[2/2] Writing to target format...\n")
 	fmt.Fprintf(os.Stderr, "      Format: %s\n", convertTargetType)
@@ -408,23 +428,12 @@ func writeDatabase(db *models.Database, dbType, outputPath, packageName, schemaF
 
 	writerOpts := newWriterOptions(outputPath, packageName, flattenSchema, nullableTypes, nullableArrays, continueOnError)
 	if extraFields != "" {
-		if !strings.EqualFold(dbType, "bun") {
+		extraFieldsJSON, err := loadExtraFields(dbType, extraFields)
-			return fmt.Errorf("--extra-fields is only supported for Bun output")
-		}
-		extraFieldsJSON, err := os.ReadFile(extraFields)
 		if err != nil {
-			return fmt.Errorf("failed to read --extra-fields file %q: %w", extraFields, err)
+			return err
-		}
-
-		var parsed []wbun.ExtraFieldConfig
-		if err := stdjson.Unmarshal(extraFieldsJSON, &parsed); err != nil {
-			return fmt.Errorf("invalid --extra-fields JSON in %q: %w", extraFields, err)
-		}
-		if len(parsed) == 0 {
-			return fmt.Errorf("--extra-fields must contain at least one field")
 		}
 		writerOpts.Metadata = map[string]interface{}{
-			"extra_fields": string(extraFieldsJSON),
+			"extra_fields": extraFieldsJSON,
 		}
 	}
 
@@ -529,6 +538,95 @@ func writeDatabase(db *models.Database, dbType, outputPath, packageName, schemaF
 	return nil
 }
 
+// loadExtraFields reads and validates the --extra-fields JSON file, returning
+// its raw content.
+func loadExtraFields(dbType, path string) (string, error) {
+	if !strings.EqualFold(dbType, "bun") {
+		return "", fmt.Errorf("--extra-fields is only supported for Bun output")
+	}
+	extraFieldsJSON, err := os.ReadFile(path)
+	if err != nil {
+		return "", fmt.Errorf("failed to read --extra-fields file %q: %w", path, err)
+	}
+
+	var parsed []wbun.ExtraFieldConfig
+	if err := stdjson.Unmarshal(extraFieldsJSON, &parsed); err != nil {
+		return "", fmt.Errorf("invalid --extra-fields JSON in %q: %w", path, err)
+	}
+	if len(parsed) == 0 {
+		return "", fmt.Errorf("--extra-fields must contain at least one field")
+	}
+	return string(extraFieldsJSON), nil
+}
+
+// validateWriteTarget performs the checks writeDatabase makes before writing,
+// without constructing a writer or touching the output path. Used by --dry-run.
+func validateWriteTarget(db *models.Database, dbType, packageName, schemaFilter, extraFields string) error {
+	if extraFields != "" {
+		if _, err := loadExtraFields(dbType, extraFields); err != nil {
+			return err
+		}
+	}
+
+	switch strings.ToLower(dbType) {
+	case "dbml", "dctx", "drawdb", "json", "yaml", "yml", "drizzle",
+		"pgsql", "postgres", "postgresql", "sql",
+		"mssql", "sqlserver", "mssql2016", "mssql2017", "mssql2019", "mssql2022",
+		"sqlite", "sqlite3", "prisma", "typeorm", "graphql", "gql":
+	case "gorm", "bun":
+		if packageName == "" {
+			return fmt.Errorf("package name is required for %s format (use --package flag)", dbType)
+		}
+	default:
+		return fmt.Errorf("unsupported target format: %s", dbType)
+	}
+
+	if schemaFilter != "" {
+		for _, schema := range db.Schemas {
+			if schema.Name == schemaFilter {
+				return nil
+			}
+		}
+		return fmt.Errorf("schema '%s' not found in database. Available schemas: %v",
+			schemaFilter, getSchemaNames(db))
+	}
+
+	if strings.EqualFold(dbType, "dctx") {
+		if len(db.Schemas) == 0 {
+			return fmt.Errorf("no schemas found in database")
+		}
+		if len(db.Schemas) > 1 {
+			return fmt.Errorf("multiple schemas found, please specify which schema to export using --schema flag. Available schemas: %v",
+				getSchemaNames(db))
+		}
+	}
+	return nil
+}
+
+// outWriter returns the command's stdout, falling back to os.Stdout when the
+// command is nil (tests call the run functions directly).
+func outWriter(cmd *cobra.Command) io.Writer {
+	if cmd == nil {
+		return os.Stdout
+	}
+	return cmd.OutOrStdout()
+}
+
+// printDryRunPlan prints the schemas and tables that would be written.
+func printDryRunPlan(out io.Writer, db *models.Database) {
+	for _, schema := range db.Schemas {
+		names := make([]string, 0, len(schema.Tables))
+		for _, t := range schema.Tables {
+			names = append(names, t.Name)
+		}
+		fmt.Fprintf(out, "    schema %q: %d table(s)", schema.Name, len(schema.Tables))
+		if len(names) > 0 {
+			fmt.Fprintf(out, " [%s]", strings.Join(names, ", "))
+		}
+		fmt.Fprintln(out)
+	}
+}
+
 // getSchemaNames returns a slice of schema names from a database
 func getSchemaNames(db *models.Database) []string {
 	names := make([]string, len(db.Schemas))

cmd/relspec/dry_run_test.go

@@ -0,0 +1,158 @@
+package main
+
+import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+)
+
+func newDryRunCmd() (*cobra.Command, *bytes.Buffer) {
+	var buf bytes.Buffer
+	cmd := &cobra.Command{}
+	cmd.SetOut(&buf)
+	return cmd, &buf
+}
+
+func TestRunConvert_DryRunWritesNothing(t *testing.T) {
+	defer func(a, b, c, d string, e bool) {
+		convertSourceType, convertSourcePath, convertTargetType, convertTargetPath, convertDryRun = a, b, c, d, e
+	}(convertSourceType, convertSourcePath, convertTargetType, convertTargetPath, convertDryRun)
+
+	dir := t.TempDir()
+	in := filepath.Join(dir, "in.json")
+	out := filepath.Join(dir, "out.json")
+	writeTestJSON(t, in, []string{"users", "posts"})
+
+	convertSourceType, convertSourcePath = "json", in
+	convertTargetType, convertTargetPath = "json", out
+	convertDryRun = true
+
+	cmd, buf := newDryRunCmd()
+	if err := runConvert(cmd, nil); err != nil {
+		t.Fatalf("dry run: %v", err)
+	}
+	if _, err := os.Stat(out); !os.IsNotExist(err) {
+		t.Fatal("dry run must not create the output file")
+	}
+	for _, want := range []string{"dry run", "users", "posts", out} {
+		if !strings.Contains(buf.String(), want) {
+			t.Errorf("plan missing %q:\n%s", want, buf.String())
+		}
+	}
+
+	// Normal behavior is unchanged.
+	convertDryRun = false
+	if err := runConvert(cmd, nil); err != nil {
+		t.Fatalf("real run: %v", err)
+	}
+	if _, err := os.Stat(out); err != nil {
+		t.Fatalf("real run should write output: %v", err)
+	}
+}
+
+func TestRunConvert_DryRunValidatesTarget(t *testing.T) {
+	defer func(a, b, c, d string, e bool) {
+		convertSourceType, convertSourcePath, convertTargetType, convertTargetPath, convertDryRun = a, b, c, d, e
+	}(convertSourceType, convertSourcePath, convertTargetType, convertTargetPath, convertDryRun)
+
+	dir := t.TempDir()
+	in := filepath.Join(dir, "in.json")
+	out := filepath.Join(dir, "models")
+	writeTestJSON(t, in, []string{"users"})
+
+	convertSourceType, convertSourcePath = "json", in
+	convertDryRun = true
+
+	// gorm without --package must fail validation, as a real run would.
+	convertTargetType, convertTargetPath = "gorm", out
+	cmd, _ := newDryRunCmd()
+	err := runConvert(cmd, nil)
+	if err == nil || !strings.Contains(err.Error(), "package name is required") {
+		t.Fatalf("expected package validation error, got %v", err)
+	}
+
+	convertTargetType = "nope"
+	err = runConvert(cmd, nil)
+	if err == nil || !strings.Contains(err.Error(), "unsupported target format") {
+		t.Fatalf("expected unsupported format error, got %v", err)
+	}
+	if _, statErr := os.Stat(out); !os.IsNotExist(statErr) {
+		t.Fatal("dry run must not create the output path")
+	}
+}
+
+func TestRunSplit_DryRunWritesNothing(t *testing.T) {
+	defer func(a, b, c, d, e string, f bool) {
+		splitSourceType, splitSourcePath, splitTargetType, splitTargetPath, splitTables, splitDryRun = a, b, c, d, e, f
+	}(splitSourceType, splitSourcePath, splitTargetType, splitTargetPath, splitTables, splitDryRun)
+
+	dir := t.TempDir()
+	in := filepath.Join(dir, "in.json")
+	out := filepath.Join(dir, "subset.json")
+	writeTestJSON(t, in, []string{"users", "posts", "comments"})
+
+	splitSourceType, splitSourcePath = "json", in
+	splitTargetType, splitTargetPath = "json", out
+	splitTables = "users,posts"
+	splitDryRun = true
+
+	cmd, buf := newDryRunCmd()
+	if err := runSplit(cmd, nil); err != nil {
+		t.Fatalf("dry run: %v", err)
+	}
+	if _, err := os.Stat(out); !os.IsNotExist(err) {
+		t.Fatal("dry run must not create the output file")
+	}
+	got := buf.String()
+	if !strings.Contains(got, "2 table(s)") || strings.Contains(got, "comments") {
+		t.Errorf("plan should show only the 2 selected tables:\n%s", got)
+	}
+
+	// A selection that matches nothing fails validation in dry-run too.
+	splitTables = "does_not_exist"
+	if err := runSplit(cmd, nil); err == nil {
+		t.Fatal("expected error for empty selection")
+	}
+}
+
+func TestRunMerge_DryRunWritesNothing(t *testing.T) {
+	saved := saveMergeState()
+	defer restoreMergeState(saved)
+	defer func(v bool) { mergeDryRun = v }(mergeDryRun)
+
+	dir := t.TempDir()
+	target := filepath.Join(dir, "target.json")
+	source := filepath.Join(dir, "source.json")
+	out := filepath.Join(dir, "merged.json")
+	writeTestJSON(t, target, []string{"users"})
+	writeTestJSON(t, source, []string{"posts"})
+
+	mergeTargetType, mergeTargetPath, mergeTargetConn = "json", target, ""
+	mergeSourceType, mergeSourcePath, mergeSourceConn = "json", source, ""
+	mergeFromList = nil
+	mergeOutputType, mergeOutputPath, mergeOutputConn = "json", out, ""
+	mergeSkipTables, mergeReportPath = "", ""
+	mergeDryRun = true
+
+	cmd, buf := newDryRunCmd()
+	if err := runMerge(cmd, nil); err != nil {
+		t.Fatalf("dry run: %v", err)
+	}
+	if _, err := os.Stat(out); !os.IsNotExist(err) {
+		t.Fatal("dry run must not create the output file")
+	}
+	for _, want := range []string{"dry run", "users", "posts", out} {
+		if !strings.Contains(buf.String(), want) {
+			t.Errorf("plan missing %q:\n%s", want, buf.String())
+		}
+	}
+
+	mergeOutputType = "nope"
+	if err := runMerge(cmd, nil); err == nil || !strings.Contains(err.Error(), "unsupported format") {
+		t.Fatalf("expected unsupported output format error, got %v", err)
+	}
+}

cmd/relspec/merge.go

@@ -61,6 +61,7 @@ var (
 	mergeReportPath    string // Path to write merge report
 	mergeFullDDL       bool   // Execute full DDL instead of diffing the live pgsql database
 	mergeFlattenSchema bool
+	mergeDryRun        bool
 )
 
 var mergeCmd = &cobra.Command{
@@ -130,6 +131,7 @@ func init() {
 	mergeCmd.Flags().BoolVar(&mergeVerbose, "verbose", false, "Show verbose output")
 	mergeCmd.Flags().BoolVar(&mergeFullDDL, "full-ddl", false, "pgsql database output: execute the full DDL instead of only the differences from the live database")
 	mergeCmd.Flags().StringVar(&mergeReportPath, "merge-report", "", "Path to write merge report (JSON format)")
+	mergeCmd.Flags().BoolVar(&mergeDryRun, "dry-run", false, "Read and merge in memory, then print the merge plan without writing any output")
 	mergeCmd.Flags().BoolVar(&mergeFlattenSchema, "flatten-schema", false, "Flatten schema.table names to schema_table (useful for databases like SQLite that do not support schemas)")
 }
 
@@ -264,6 +266,29 @@ func runMerge(cmd *cobra.Command, args []string) error {
 			merge.GetColumnTypeConflictSummary(result, 10))
 	}
 
+	if mergeDryRun {
+		if !isMergeOutputFormat(mergeOutputType) {
+			return fmt.Errorf("dry run validation failed: Output: unsupported format '%s'", mergeOutputType)
+		}
+		out := outWriter(cmd)
+		fmt.Fprintf(out, "RelSpec merge plan (dry run - nothing written):\n")
+		fmt.Fprintf(out, "  Target: %s database '%s'\n", mergeTargetType, targetDB.Name)
+		fmt.Fprintf(out, "  Source: %s database '%s'\n", mergeSourceType, sourceDB.Name)
+		switch {
+		case mergeOutputPath != "":
+			fmt.Fprintf(out, "  Output: %s -> %s\n", mergeOutputType, mergeOutputPath)
+		case mergeOutputConn != "":
+			fmt.Fprintf(out, "  Output: %s -> %s\n", mergeOutputType, maskPassword(mergeOutputConn))
+		default:
+			fmt.Fprintf(out, "  Output: %s\n", mergeOutputType)
+		}
+		fmt.Fprintf(out, "  Result:\n")
+		printDryRunPlan(out, targetDB)
+		fmt.Fprintf(out, "\n%s\n", merge.GetMergeSummary(result))
+		fmt.Fprintf(os.Stderr, "\n=== Dry run complete: no output written ===\n")
+		return nil
+	}
+
 	// Step 4: Write output
 	fmt.Fprintf(os.Stderr, "\n[4/4] Writing output...\n")
 	fmt.Fprintf(os.Stderr, "      Format: %s\n", mergeOutputType)
@@ -285,6 +310,16 @@ func runMerge(cmd *cobra.Command, args []string) error {
 	return nil
 }
 
+// isMergeOutputFormat reports whether writeDatabaseForMerge supports dbType.
+func isMergeOutputFormat(dbType string) bool {
+	switch strings.ToLower(dbType) {
+	case "dbml", "dctx", "drawdb", "graphql", "json", "yaml", "gorm", "bun",
+		"drizzle", "prisma", "typeorm", "sqlite", "sqlite3", "pgsql":
+		return true
+	}
+	return false
+}
+
 func readDatabaseForMerge(dbType, filePath, connString, label string) (*models.Database, error) {
 	var reader readers.Reader
 

cmd/relspec/split.go

@@ -24,6 +24,7 @@ var (
 	splitExcludeTables  string
 	splitNullableTypes  string
 	splitNullableArrays string
+	splitDryRun         bool
 )
 
 var splitCmd = &cobra.Command{
@@ -115,6 +116,8 @@ func init() {
 	splitCmd.Flags().StringVar(&splitNullableTypes, "types", "", "Nullable type package for code-gen writers (bun/gorm): 'baselib' (default, Go pointer types), 'stdlib' (database/sql), or 'sqltypes'")
 	splitCmd.Flags().StringVar(&splitNullableArrays, "array-nullable", "", "Nullable PostgreSQL array representation for the Bun writer in stdlib/baselib --types mode: 'slice' (default, plain slice) or 'pointer_slice' (*[]T, distinguishes NULL from '{}')")
 
+	splitCmd.Flags().BoolVar(&splitDryRun, "dry-run", false, "Read, filter and validate the selection and print the plan without writing any output")
+
 	err := splitCmd.MarkFlagRequired("from")
 	if err != nil {
 		fmt.Fprintf(os.Stderr, "Error marking from flag as required: %v\n", err)
@@ -174,6 +177,26 @@ func runSplit(cmd *cobra.Command, args []string) error {
 	}
 	fmt.Fprintf(os.Stderr, "      ✓ Filtered to: %d schema(s), %d table(s)\n\n", len(filteredDB.Schemas), filteredTables)
 
+	if splitDryRun {
+		if err := validateWriteTarget(filteredDB, splitTargetType, splitPackageName, "", ""); err != nil {
+			return fmt.Errorf("dry run validation failed: %w", err)
+		}
+		out := outWriter(cmd)
+		fmt.Fprintf(out, "RelSpec split plan (dry run - nothing written):\n")
+		fmt.Fprintf(out, "  Input:  %s database '%s'\n", splitSourceType, db.Name)
+		fmt.Fprintf(out, "  Output: %s -> %s\n", splitTargetType, splitTargetPath)
+		fmt.Fprintf(out, "  Selection: %s\n", splitSelection{
+			Schemas:        parseCommaSeparated(splitSchemas),
+			Tables:         parseCommaSeparated(splitTables),
+			ExcludeSchemas: parseCommaSeparated(splitExcludeSchema),
+			ExcludeTables:  parseCommaSeparated(splitExcludeTables),
+			DatabaseName:   splitDatabaseName,
+		}.summary())
+		printDryRunPlan(out, filteredDB)
+		fmt.Fprintf(os.Stderr, "=== Dry run complete: no output written ===\n\n")
+		return nil
+	}
+
 	// Write to target format
 	fmt.Fprintf(os.Stderr, "[3/3] Writing to target format...\n")
 	fmt.Fprintf(os.Stderr, "      Format: %s\n", splitTargetType)