validation/validator.go

package validation

import (
	"fmt"
	"net/url"
	"regexp"
	"strings"

	"slices"

	"github.com/korap/korap-mcp/service"
	"github.com/rs/zerolog"
)

// Validator provides input validation and response schema validation
type Validator struct {
	logger zerolog.Logger
}

// New creates a new validator instance
func New(logger zerolog.Logger) *Validator {
	return &Validator{
		logger: logger.With().Str("component", "validator").Logger(),
	}
}

// SearchRequest holds the parameters for a search request validation
type SearchRequest struct {
	Query         string `json:"query"`
	QueryLanguage string `json:"query_language,omitempty"`
	Corpus        string `json:"corpus,omitempty"`
	Count         int    `json:"count,omitempty"`
}

// MetadataRequest holds the parameters for a metadata request validation
type MetadataRequest struct {
	Action string `json:"action"`
	Corpus string `json:"corpus,omitempty"`
}

// ValidationError represents a validation error with details
type ValidationError struct {
	Field   string `json:"field"`
	Value   string `json:"value"`
	Message string `json:"message"`
}

func (e ValidationError) Error() string {
	return fmt.Sprintf("validation error for field '%s' (value: '%s'): %s", e.Field, e.Value, e.Message)
}

// ValidationErrors represents multiple validation errors
type ValidationErrors struct {
	Errors []ValidationError `json:"errors"`
}

func (e ValidationErrors) Error() string {
	if len(e.Errors) == 0 {
		return "validation errors occurred"
	}
	var messages []string
	for _, err := range e.Errors {
		messages = append(messages, err.Error())
	}
	return strings.Join(messages, "; ")
}

// Regular expressions for validation
var (
	// Query language validation
	validQueryLanguages = []string{"poliqarp", "poliqarpplus", "cosmas2", "annis", "cql", "cqp", "fcsql"}

	// Corpus ID validation - KorAP collection queries with metadata fields, operators, and regex
	corpusIDRegex = regexp.MustCompile(`^[a-zA-Z0-9._\-\s&|!=<>()/*"']+$`)

	// Action validation for metadata requests
	validMetadataActions = map[string]bool{
		"list":       true,
		"statistics": true,
	}
)

// ValidateSearchRequest validates a search request
func (v *Validator) ValidateSearchRequest(req SearchRequest) error {
	var errors []ValidationError

	// Validate query - required and non-empty
	if strings.TrimSpace(req.Query) == "" {
		errors = append(errors, ValidationError{
			Field:   "query",
			Value:   req.Query,
			Message: "query is required and cannot be empty",
		})
	} else {
		// Basic query validation - check for potentially dangerous patterns
		if err := v.validateQuerySafety(req.Query); err != nil {
			errors = append(errors, ValidationError{
				Field:   "query",
				Value:   req.Query,
				Message: err.Error(),
			})
		}
	}

	// Validate query language if provided
	if req.QueryLanguage != "" && !slices.Contains(validQueryLanguages, req.QueryLanguage) {
		errors = append(errors, ValidationError{
			Field:   "query_language",
			Value:   req.QueryLanguage,
			Message: fmt.Sprintf("invalid query language, must be one of: %s", strings.Join(validQueryLanguages, ", ")),
		})
	}

	// Validate corpus if provided
	if req.Corpus != "" {
		if err := v.validateCorpusID(req.Corpus); err != nil {
			errors = append(errors, ValidationError{
				Field:   "corpus",
				Value:   req.Corpus,
				Message: err.Error(),
			})
		}
	}

	// Validate count if provided (0 means use default, so only validate non-zero values)
	if req.Count < 0 || req.Count > 10000 {
		errors = append(errors, ValidationError{
			Field:   "count",
			Value:   fmt.Sprintf("%d", req.Count),
			Message: "count must be between 0 and 10000 (0 means use default)",
		})
	}

	if len(errors) > 0 {
		v.logger.Warn().Interface("errors", errors).Msg("Search request validation failed")
		return ValidationErrors{Errors: errors}
	}

	v.logger.Debug().Interface("request", req).Msg("Search request validation passed")
	return nil
}

// ValidateMetadataRequest validates a metadata request
func (v *Validator) ValidateMetadataRequest(req MetadataRequest) error {
	var errors []ValidationError

	// Validate action - required
	if strings.TrimSpace(req.Action) == "" {
		errors = append(errors, ValidationError{
			Field:   "action",
			Value:   req.Action,
			Message: "action is required and cannot be empty",
		})
	} else if !validMetadataActions[req.Action] {
		var validActions []string
		for action := range validMetadataActions {
			validActions = append(validActions, action)
		}
		errors = append(errors, ValidationError{
			Field:   "action",
			Value:   req.Action,
			Message: fmt.Sprintf("invalid action, must be one of: %s", strings.Join(validActions, ", ")),
		})
	}

	// Validate corpus if provided
	if req.Corpus != "" {
		if err := v.validateCorpusID(req.Corpus); err != nil {
			errors = append(errors, ValidationError{
				Field:   "corpus",
				Value:   req.Corpus,
				Message: err.Error(),
			})
		}
	}

	if len(errors) > 0 {
		v.logger.Warn().Interface("errors", errors).Msg("Metadata request validation failed")
		return ValidationErrors{Errors: errors}
	}

	v.logger.Debug().Interface("request", req).Msg("Metadata request validation passed")
	return nil
}

// ValidateSearchResponse validates a search response
func (v *Validator) ValidateSearchResponse(resp *service.SearchResponse) error {
	if resp == nil {
		return fmt.Errorf("search response is nil")
	}

	var errors []ValidationError

	// Validate meta structure
	if resp.Meta.TotalResults < 0 {
		errors = append(errors, ValidationError{
			Field:   "meta.totalResults",
			Value:   fmt.Sprintf("%d", resp.Meta.TotalResults),
			Message: "totalResults cannot be negative",
		})
	}

	if resp.Meta.Count < 0 {
		errors = append(errors, ValidationError{
			Field:   "meta.count",
			Value:   fmt.Sprintf("%d", resp.Meta.Count),
			Message: "count cannot be negative",
		})
	}

	if resp.Meta.StartIndex < 0 {
		errors = append(errors, ValidationError{
			Field:   "meta.startIndex",
			Value:   fmt.Sprintf("%d", resp.Meta.StartIndex),
			Message: "startIndex cannot be negative",
		})
	}

	if resp.Meta.ItemsPerPage < 0 {
		errors = append(errors, ValidationError{
			Field:   "meta.itemsPerPage",
			Value:   fmt.Sprintf("%d", resp.Meta.ItemsPerPage),
			Message: "itemsPerPage cannot be negative",
		})
	}

	// Validate matches if present
	if resp.Matches != nil {
		for i, match := range resp.Matches {
			if match.MatchID == "" {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("matches[%d].matchID", i),
					Value:   "",
					Message: "match ID is required",
				})
			}

			if match.TextSigle == "" {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("matches[%d].textSigle", i),
					Value:   "",
					Message: "textSigle is required",
				})
			}

			if match.Position < 0 {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("matches[%d].position", i),
					Value:   fmt.Sprintf("%d", match.Position),
					Message: "position cannot be negative",
				})
			}
		}
	}

	if len(errors) > 0 {
		v.logger.Warn().Interface("errors", errors).Msg("Search response validation failed")
		return ValidationErrors{Errors: errors}
	}

	v.logger.Debug().Msg("Search response validation passed")
	return nil
}

// ValidateCorpusListResponse validates a corpus list response
func (v *Validator) ValidateCorpusListResponse(resp *service.CorpusListResponse) error {
	if resp == nil {
		return fmt.Errorf("corpus list response is nil")
	}

	var errors []ValidationError

	// Validate corpus entries
	if resp.Corpora != nil {
		for i, corpus := range resp.Corpora {
			if corpus.ID == "" {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("corpora[%d].id", i),
					Value:   "",
					Message: "corpus ID is required",
				})
			} else if err := v.validateCorpusID(corpus.ID); err != nil {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("corpora[%d].id", i),
					Value:   corpus.ID,
					Message: err.Error(),
				})
			}

			if corpus.Name == "" {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("corpora[%d].name", i),
					Value:   "",
					Message: "corpus name is required",
				})
			}

			if corpus.Documents < 0 {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("corpora[%d].documents", i),
					Value:   fmt.Sprintf("%d", corpus.Documents),
					Message: "document count cannot be negative",
				})
			}

			if corpus.Tokens < 0 {
				errors = append(errors, ValidationError{
					Field:   fmt.Sprintf("corpora[%d].tokens", i),
					Value:   fmt.Sprintf("%d", corpus.Tokens),
					Message: "token count cannot be negative",
				})
			}
		}
	}

	if len(errors) > 0 {
		v.logger.Warn().Interface("errors", errors).Msg("Corpus list response validation failed")
		return ValidationErrors{Errors: errors}
	}

	v.logger.Debug().Msg("Corpus list response validation passed")
	return nil
}

// ValidateStatisticsResponse validates a statistics response
func (v *Validator) ValidateStatisticsResponse(resp *service.StatisticsResponse) error {
	if resp == nil {
		return fmt.Errorf("statistics response is nil")
	}

	var errors []ValidationError

	if resp.Documents < 0 {
		errors = append(errors, ValidationError{
			Field:   "documents",
			Value:   fmt.Sprintf("%d", resp.Documents),
			Message: "document count cannot be negative",
		})
	}

	if resp.Tokens < 0 {
		errors = append(errors, ValidationError{
			Field:   "tokens",
			Value:   fmt.Sprintf("%d", resp.Tokens),
			Message: "token count cannot be negative",
		})
	}

	if resp.Sentences < 0 {
		errors = append(errors, ValidationError{
			Field:   "sentences",
			Value:   fmt.Sprintf("%d", resp.Sentences),
			Message: "sentence count cannot be negative",
		})
	}

	if resp.Paragraphs < 0 {
		errors = append(errors, ValidationError{
			Field:   "paragraphs",
			Value:   fmt.Sprintf("%d", resp.Paragraphs),
			Message: "paragraph count cannot be negative",
		})
	}

	if len(errors) > 0 {
		v.logger.Warn().Interface("errors", errors).Msg("Statistics response validation failed")
		return ValidationErrors{Errors: errors}
	}

	v.logger.Debug().Msg("Statistics response validation passed")
	return nil
}

// validateQuerySafety performs basic security validation on queries
func (v *Validator) validateQuerySafety(query string) error {
	// Check for extremely long queries that could cause DoS
	if len(query) > 10000 {
		return fmt.Errorf("query is too long (max 10000 characters)")
	}

	// Check for potentially dangerous URL patterns
	if strings.Contains(query, "://") {
		if _, err := url.Parse(query); err == nil {
			return fmt.Errorf("query appears to contain a URL which is not allowed")
		}
	}

	// Check for excessive nesting that could cause parser issues
	openParens := strings.Count(query, "(")
	closeParens := strings.Count(query, ")")
	if openParens != closeParens {
		return fmt.Errorf("unmatched parentheses in query")
	}
	if openParens > 100 {
		return fmt.Errorf("query has too many nested levels (max 100)")
	}

	return nil
}

// validateCorpusID validates a corpus identifier or collection query
// This supports both simple corpus sigles (e.g., "DeReKo-2023-I") and complex
// collection queries with metadata fields (e.g., "textClass = \"politics\" & pubDate in 2020")
func (v *Validator) validateCorpusID(corpusID string) error {
	if len(corpusID) == 0 {
		return fmt.Errorf("corpus ID cannot be empty")
	}

	if len(corpusID) > 100 {
		return fmt.Errorf("corpus ID is too long (max 100 characters)")
	}

	if !corpusIDRegex.MatchString(corpusID) {
		return fmt.Errorf("collection query contains invalid characters (supports alphanumeric, dots, hyphens, underscores, spaces, quotes, operators & | ! = < > in, parentheses, and regex /pattern/)")
	}

	return nil
}

// SanitizeQuery performs basic sanitization on search queries
func (v *Validator) SanitizeQuery(query string) string {
	// Trim whitespace
	sanitized := strings.TrimSpace(query)

	// Remove any null bytes
	sanitized = strings.ReplaceAll(sanitized, "\x00", "")

	// Normalize whitespace
	sanitized = regexp.MustCompile(`\s+`).ReplaceAllString(sanitized, " ")

	v.logger.Debug().
		Str("original", query).
		Str("sanitized", sanitized).
		Msg("Query sanitized")

	return sanitized
}

// SanitizeCorpusID performs basic sanitization on corpus IDs
func (v *Validator) SanitizeCorpusID(corpusID string) string {
	// Trim whitespace
	sanitized := strings.TrimSpace(corpusID)

	// Remove any null bytes
	sanitized = strings.ReplaceAll(sanitized, "\x00", "")

	// Convert to lowercase for consistency
	sanitized = strings.ToLower(sanitized)

	v.logger.Debug().
		Str("original", corpusID).
		Str("sanitized", sanitized).
		Msg("Corpus ID sanitized")

	return sanitized
}