relspecgo/pkg/writers/writer.go

package writers

import (
	"regexp"
	"strings"

	"git.warky.dev/wdevs/relspecgo/pkg/models"
)

// Writer defines the interface for writing database specifications
// to various output formats at different granularity levels
type Writer interface {
	// WriteDatabase takes a Database model and writes it to the desired format
	WriteDatabase(db *models.Database) error

	// WriteSchema takes a Schema model and writes it to the desired format
	WriteSchema(schema *models.Schema) error

	// WriteTable takes a Table model and writes it to the desired format
	WriteTable(table *models.Table) error
}

// WriterOptions contains common options for writers
type WriterOptions struct {
	// OutputPath is the path where the output should be written
	OutputPath string

	// PackageName is the Go package name (for code generation)
	PackageName string

	// FlattenSchema disables schema.table dot notation and instead joins
	// schema and table with an underscore (e.g., "public_users").
	// Useful for databases like SQLite that do not support schemas.
	FlattenSchema bool

	// Additional options can be added here as needed
	Metadata map[string]interface{}
}

// QualifiedTableName returns a schema-qualified table name.
// When flatten is true, schema and table are joined with underscore (e.g., "schema_table").
// When flatten is false, they are dot-separated (e.g., "schema.table").
// If schema is empty, just the table name is returned regardless of flatten.
func QualifiedTableName(schema, table string, flatten bool) string {
	if schema == "" {
		return table
	}
	if flatten {
		return schema + "_" + table
	}
	return schema + "." + table
}

// SanitizeFilename removes quotes, comments, and invalid characters from identifiers
// to make them safe for use in filenames. This handles:
// - Double and single quotes: "table_name" or 'table_name' -> table_name
// - DBML comments: table [note: 'description'] -> table
// - Invalid filename characters: replaced with underscores
func SanitizeFilename(name string) string {
	// Remove DBML/DCTX style comments in brackets (e.g., [note: 'description'])
	commentRegex := regexp.MustCompile(`\s*\[.*?\]\s*`)
	name = commentRegex.ReplaceAllString(name, "")

	// Remove quotes (both single and double)
	name = strings.ReplaceAll(name, `"`, "")
	name = strings.ReplaceAll(name, `'`, "")

	// Remove backticks (MySQL style identifiers)
	name = strings.ReplaceAll(name, "`", "")

	// Replace invalid filename characters with underscores
	// Invalid chars: / \ : * ? " < > | and control characters
	invalidChars := regexp.MustCompile(`[/\\:*?"<>|\x00-\x1f\x7f]`)
	name = invalidChars.ReplaceAllString(name, "_")

	// Trim whitespace and consecutive underscores
	name = strings.TrimSpace(name)
	name = regexp.MustCompile(`_+`).ReplaceAllString(name, "_")
	name = strings.Trim(name, "_")

	return name
}

// QuoteDefaultValue wraps a sanitized default value in single quotes when the SQL
// column type requires it (strings, dates, times, UUIDs, enums). Numeric types
// (integers, floats, serials) and boolean types are left unquoted. Function-call
// expressions such as now() or gen_random_uuid() are always left unquoted regardless
// of type, because they contain parentheses.
//
// Examples (varchar):   "disconnected" → "'disconnected'"
// Examples (boolean):   "true"         → "true"
// Examples (bigint):    "0"            → "0"
// Examples (timestamp): "now()"        → "now()"   (function call – never quoted)
func QuoteDefaultValue(value, sqlType string) string {
	// Function calls are never quoted regardless of column type.
	if strings.Contains(value, "(") || strings.Contains(value, ")") {
		return value
	}

	// Normalise the SQL type: lowercase, strip length/precision suffix.
	baseType := strings.ToLower(strings.TrimSpace(sqlType))
	if idx := strings.Index(baseType, "("); idx > 0 {
		baseType = baseType[:idx]
	}

	// Types whose default values must NOT be quoted.
	unquotedTypes := map[string]bool{
		// Integer types
		"integer":     true,
		"int":         true,
		"int2":        true,
		"int4":        true,
		"int8":        true,
		"smallint":    true,
		"bigint":      true,
		"serial":      true,
		"smallserial": true,
		"bigserial":   true,
		// Float / numeric types
		"real":             true,
		"float":            true,
		"float4":           true,
		"float8":           true,
		"double precision": true,
		"numeric":          true,
		"decimal":          true,
		"money":            true,
		// Boolean
		"boolean": true,
		"bool":    true,
	}

	if unquotedTypes[baseType] {
		return value
	}

	// Everything else (text, varchar, char, uuid, date, time, timestamp, json, …)
	// is treated as a quoted literal.
	return "'" + value + "'"
}

// SanitizeStructTagValue sanitizes a value to be safely used inside Go struct tags.
// Go struct tags are delimited by backticks, so any backtick in the value would break the syntax.
// This function:
// - Removes DBML/DCTX comments in brackets
// - Removes all quotes (double, single, and backticks)
// - Returns a clean identifier safe for use in struct tags and field names
func SanitizeStructTagValue(value string) string {
	// Remove DBML/DCTX style comments in brackets (e.g., [note: 'description'])
	commentRegex := regexp.MustCompile(`\s*\[.*?\]\s*`)
	value = commentRegex.ReplaceAllString(value, "")

	// Trim whitespace
	value = strings.TrimSpace(value)

	// Remove all quotes: backticks, double quotes, and single quotes
	// This ensures the value is clean for use as Go identifiers and struct tag values
	value = strings.ReplaceAll(value, "`", "")
	value = strings.ReplaceAll(value, `"`, "")
	value = strings.ReplaceAll(value, `'`, "")

	return value
}