BigChef/processor/regex.go

package processor

import (
	"cook/utils"
	"fmt"
	"regexp"
	"strconv"
	"strings"
	"time"

	logger "git.site.quack-lab.dev/dave/cylogger"
	lua "github.com/yuin/gopher-lua"
)

// regexLogger is a scoped logger for the processor/regex package.
var regexLogger = logger.Default.WithPrefix("processor/regex")

type CaptureGroup struct {
	Name    string
	Value   string
	Updated string
	Range   [2]int
}

// ProcessContent applies regex replacement with Lua processing
// The filename here exists ONLY so we can pass it to the lua environment
// It's not used for anything else
func ProcessRegex(content string, command utils.ModifyCommand, filename string) ([]utils.ReplaceCommand, error) {
	processRegexLogger := regexLogger.WithPrefix("ProcessRegex").WithField("commandName", command.Name).WithField("file", filename)
	processRegexLogger.Debug("Starting regex processing for file")
	processRegexLogger.Trace("Initial file content length: %d", len(content))
	processRegexLogger.Trace("Command details: %+v", command)

	var commands []utils.ReplaceCommand
	// Start timing the regex processing
	startTime := time.Now()

	// We don't HAVE to do this multiple times for a pattern
	// But it's quick enough for us to not care
	pattern := resolveRegexPlaceholders(command.Regex)
	processRegexLogger.Debug("Resolved regex placeholders. Pattern: %s", pattern)

	// I'm not too happy about having to trim regex, we could have meaningful whitespace or newlines
	// But it's a compromise that allows us to use | in yaml
	// Otherwise we would have to escape every god damn pair of quotation marks
	// And a bunch of other shit
	pattern = strings.TrimSpace(pattern)
	processRegexLogger.Debug("Trimmed regex pattern: %s", pattern)

	patternCompileStart := time.Now()
	compiledPattern, err := regexp.Compile(pattern)
	if err != nil {
		processRegexLogger.Error("Error compiling pattern %q: %v", pattern, err)
		return commands, fmt.Errorf("error compiling pattern: %v", err)
	}
	processRegexLogger.Debug("Compiled pattern successfully in %v", time.Since(patternCompileStart))

	// Same here, it's just string concatenation, it won't kill us
	// More important is that we don't fuck up the command
	// But we shouldn't be able to since it's passed by value
	previousLuaExpr := command.Lua
	luaExpr := BuildLuaScript(command.Lua)
	processRegexLogger.Debug("Transformed Lua expression: %q → %q", previousLuaExpr, luaExpr)
	processRegexLogger.Trace("Full Lua script: %q", utils.LimitString(luaExpr, 200))

	// Process all regex matches
	matchFindStart := time.Now()
	indices := compiledPattern.FindAllStringSubmatchIndex(content, -1)
	matchFindDuration := time.Since(matchFindStart)

	processRegexLogger.Debug("Found %d matches in content of length %d (search took %v)",
		len(indices), len(content), matchFindDuration)
	processRegexLogger.Trace("Match indices: %v", indices)

	// Log pattern complexity metrics
	patternComplexity := estimatePatternComplexity(pattern)
	processRegexLogger.Debug("Pattern complexity estimate: %d", patternComplexity)

	if len(indices) == 0 {
		processRegexLogger.Warning("No matches found for regex: %q", pattern)
		processRegexLogger.Debug("Total regex processing time: %v", time.Since(startTime))
		return commands, nil
	}

	// We walk backwards because we're replacing something with something else that might be longer
	// And in the case it is longer than the original all indicces past that change will be fucked up
	// By going backwards we fuck up all the indices to the end of the file that we don't care about
	// Because there either aren't any (last match) or they're already modified (subsequent matches)
	for i, matchIndices := range indices {
		matchLogger := processRegexLogger.WithField("matchNum", i+1)
		matchLogger.Debug("Processing match %d of %d", i+1, len(indices))
		matchLogger.Trace("Match indices: %v (match position %d-%d)", matchIndices, matchIndices[0], matchIndices[1])

		L, err := NewLuaState()
		if err != nil {
			matchLogger.Error("Error creating Lua state: %v", err)
			return commands, fmt.Errorf("error creating Lua state: %v", err)
		}
		L.SetGlobal("file", lua.LString(filename))
		// Hmm... Maybe we don't want to defer this..
		// Maybe we want to close them every iteration
		// We'll leave it as is for now
		defer L.Close()
		matchLogger.Trace("Lua state created successfully for match %d", i+1)

		// Why we're doing this whole song and dance of indices is to properly handle empty matches
		// Plus it's a little cleaner to surgically replace our matches
		// If we were to use string.replace and encountered an empty match there'd be nothing to replace
		// But using indices an empty match would have its starting and ending indices be the same
		// So when we're cutting open the array we say 0:7 + modified + 7:end
		// As if concatenating in the middle of the array
		// Plus it supports lookarounds
		matchContent := content[matchIndices[0]:matchIndices[1]]
		matchPreview := utils.LimitString(matchContent, 50)
		matchLogger.Trace("Matched content: %q (length: %d)", matchPreview, len(matchContent))

		groups := matchIndices[2:]
		if len(groups) <= 0 {
			matchLogger.Warning("No capture groups found for match %q and regex %q", matchPreview, pattern)
			continue
		}
		if len(groups)%2 == 1 {
			matchLogger.Warning("Invalid number of group indices (%d), should be even: %v", len(groups), groups)
			continue
		}

		// Count how many valid groups we have
		validGroups := 0
		for j := 0; j < len(groups); j += 2 {
			if groups[j] != -1 && groups[j+1] != -1 {
				validGroups++
			}
		}
		matchLogger.Debug("Found %d valid capture groups in match", validGroups)

		for _, index := range groups {
			if index == -1 {
				matchLogger.Warning("Negative index encountered in match indices %v. This may indicate an issue with the regex pattern or an empty/optional capture group.", matchIndices)
				continue
			}
		}

		// We have to use array to preserve order
		// Very important for the reconstruction step
		// Because we must overwrite the values in reverse order
		// See comments a few dozen lines above for more details
		captureGroups := make([]*CaptureGroup, 0, len(groups)/2)
		groupNames := compiledPattern.SubexpNames()[1:]
		for i, name := range groupNames {
			start := groups[i*2]
			end := groups[i*2+1]
			if start == -1 || end == -1 {
				matchLogger.Debug("Skipping empty or unmatched capture group #%d (name: %q)", i+1, name)
				continue
			}

			value := content[start:end]
			captureGroups = append(captureGroups, &CaptureGroup{
				Name:  name,
				Value: value,
				Range: [2]int{start, end},
			})

			// Include name info in log if available
			if name != "" {
				matchLogger.Trace("Capture group '%s': %q (pos %d-%d)", name, value, start, end)
			} else {
				matchLogger.Trace("Capture group #%d: %q (pos %d-%d)", i+1, value, start, end)
			}
		}

		// Use the DeduplicateGroups flag to control whether to deduplicate capture groups
		if !command.NoDedup {
			matchLogger.Debug("Deduplicating capture groups as specified in command settings")
			captureGroups = deduplicateGroups(captureGroups)
			matchLogger.Trace("Capture groups after deduplication: %v", captureGroups)
		} else {
			matchLogger.Debug("Skipping deduplication of capture groups (NoDedup is true)")
		}

		if err := toLua(L, captureGroups); err != nil {
			matchLogger.Error("Failed to set Lua variables for capture groups: %v", err)
			continue
		}
		matchLogger.Debug("Set %d capture groups as Lua variables", len(captureGroups))
		matchLogger.Trace("Lua globals set for capture groups")

		if err := L.DoString(luaExpr); err != nil {
			matchLogger.Error("Lua script execution failed: %v\nScript: %s\nCapture Groups: %+v",
				err, utils.LimitString(luaExpr, 200), captureGroups)
			continue
		}
		matchLogger.Debug("Lua script executed successfully")

		// Get modifications from Lua
		updatedCaptureGroups, err := fromLua(L, captureGroups)
		if err != nil {
			matchLogger.Error("Failed to retrieve modifications from Lua: %v", err)
			continue
		}
		matchLogger.Debug("Retrieved updated values from Lua")
		matchLogger.Trace("Updated capture groups from Lua: %v", updatedCaptureGroups)

		replacement := ""
		replacementVar := L.GetGlobal("replacement")
		if replacementVar.Type() != lua.LTNil {
			replacement = replacementVar.String()
			matchLogger.Debug("Using global replacement variable from Lua: %q", replacement)
		}

		// Check if modification flag is set
		modifiedVal := L.GetGlobal("modified")
		if modifiedVal.Type() != lua.LTBool || !lua.LVAsBool(modifiedVal) {
			matchLogger.Debug("Skipping match - no modifications indicated by Lua script")
			continue
		}

		if replacement == "" {
			// Apply the modifications to the original match
			replacement = matchContent

			// Count groups that were actually modified
			modifiedGroupsCount := 0
			for _, capture := range updatedCaptureGroups {
				if capture.Value != capture.Updated {
					modifiedGroupsCount++
				}
			}
			matchLogger.Info("%d of %d capture groups identified for modification", modifiedGroupsCount, len(updatedCaptureGroups))

			for _, capture := range updatedCaptureGroups {
				if capture.Value == capture.Updated {
					matchLogger.Debug("Capture group unchanged: %s", utils.LimitString(capture.Value, 50))
					continue
				}

				// Log what changed with context
				matchLogger.Debug("Capture group %q scheduled for modification: %q → %q",
					capture.Name, utils.LimitString(capture.Value, 50), utils.LimitString(capture.Updated, 50))

				// Indices of the group are relative to content
				// To relate them to match we have to subtract the match start index
				// replacement = replacement[:groupStart] + newVal + replacement[groupEnd:]
				commands = append(commands, utils.ReplaceCommand{
					From: capture.Range[0],
					To:   capture.Range[1],
					With: capture.Updated,
				})
				matchLogger.Trace("Added replacement command: %+v", commands[len(commands)-1])
			}
		} else {
			matchLogger.Debug("Using full replacement string from Lua: %q", utils.LimitString(replacement, 50))
			commands = append(commands, utils.ReplaceCommand{
				From: matchIndices[0],
				To:   matchIndices[1],
				With: replacement,
			})
			matchLogger.Trace("Added full replacement command: %+v", commands[len(commands)-1])
		}
	}

	processRegexLogger.Debug("Total regex processing time: %v", time.Since(startTime))
	processRegexLogger.Debug("Generated %d total modifications", len(commands))
	return commands, nil
}

func deduplicateGroups(captureGroups []*CaptureGroup) []*CaptureGroup {
	deduplicateGroupsLogger := regexLogger.WithPrefix("deduplicateGroups")
	deduplicateGroupsLogger.Debug("Starting deduplication of capture groups")
	deduplicateGroupsLogger.Trace("Input capture groups: %v", captureGroups)

	// Preserve input order and drop any group that overlaps with an already accepted group
	accepted := make([]*CaptureGroup, 0, len(captureGroups))
	for _, group := range captureGroups {
		groupLogger := deduplicateGroupsLogger.WithField("groupName", group.Name).WithField("groupRange", group.Range)
		groupLogger.Debug("Processing capture group")

		overlaps := false
		for _, kept := range accepted {
			// Overlap if start < keptEnd and end > keptStart (adjacent is allowed)
			if group.Range[0] < kept.Range[1] && group.Range[1] > kept.Range[0] {
				overlaps = true
				break
			}
		}

		if overlaps {
			groupLogger.Warning("Overlapping capture group detected and skipped.")
			continue
		}

		groupLogger.Debug("Capture group does not overlap with previously accepted groups. Adding.")
		accepted = append(accepted, group)
	}

	deduplicateGroupsLogger.Debug("Finished deduplication. Original %d groups, %d deduplicated.", len(captureGroups), len(accepted))
	deduplicateGroupsLogger.Trace("Deduplicated groups: %v", accepted)

	return accepted
}

// The order of these replaces is important
// This one handles !num-s inside of named capture groups
// If it were not here our !num in a named capture group would
// Expand to another capture group in the capture group
// We really only want one (our named) capture group
func resolveRegexPlaceholders(pattern string) string {
	resolveLogger := regexLogger.WithPrefix("resolveRegexPlaceholders").WithField("originalPattern", utils.LimitString(pattern, 100))
	resolveLogger.Debug("Resolving regex placeholders in pattern")

	// Handle special pattern modifications
	if !strings.HasPrefix(pattern, "(?s)") {
		pattern = "(?s)" + pattern
		resolveLogger.Debug("Prepended '(?s)' to pattern for single-line mode")
	}

	namedGroupNum := regexp.MustCompile(`(?:(\?<[^>]+>)(!num))`)
	pattern = namedGroupNum.ReplaceAllStringFunc(pattern, func(match string) string {
		funcLogger := resolveLogger.WithPrefix("namedGroupNumReplace").WithField("match", utils.LimitString(match, 50))
		funcLogger.Debug("Processing named group !num placeholder")
		parts := namedGroupNum.FindStringSubmatch(match)
		if len(parts) != 3 {
			funcLogger.Warning("Unexpected number of submatches for namedGroupNum: %d. Returning original match.", len(parts))
			return match
		}
		replacement := `-?\d*\.?\d+`
		funcLogger.Trace("Replacing !num in named group with: %q", replacement)
		return parts[1] + replacement
	})
	resolveLogger.Debug("Handled named group !num placeholders")

	pattern = strings.ReplaceAll(pattern, "!num", `(-?\d*\.?\d+)`)
	resolveLogger.Debug("Replaced !num with numeric capture group")

	pattern = strings.ReplaceAll(pattern, "!any", `.*?`)
	resolveLogger.Debug("Replaced !any with non-greedy wildcard")

	repPattern := regexp.MustCompile(`!rep\(([^,]+),\s*(\d+)\)`)
	// !rep(pattern, count) repeats the pattern n times
	// Inserting !any between each repetition
	pattern = repPattern.ReplaceAllStringFunc(pattern, func(match string) string {
		funcLogger := resolveLogger.WithPrefix("repPatternReplace").WithField("match", utils.LimitString(match, 50))
		funcLogger.Debug("Processing !rep placeholder")
		parts := repPattern.FindStringSubmatch(match)
		if len(parts) != 3 {
			funcLogger.Warning("Unexpected number of submatches for repPattern: %d. Returning original match.", len(parts))
			return match
		}
		repeatedPattern := parts[1]
		countStr := parts[2]
		repetitions, err := strconv.Atoi(countStr)
		if err != nil {
			funcLogger.Error("Failed to parse repetition count %q: %v. Returning original match.", countStr, err)
			return match
		}

		var finalReplacement string
		if repetitions > 0 {
			finalReplacement = strings.Repeat(repeatedPattern+".*?", repetitions-1) + repeatedPattern
		} else {
			finalReplacement = ""
		}

		funcLogger.Trace("Replaced !rep with %d repetitions of %q: %q", repetitions, utils.LimitString(repeatedPattern, 30), utils.LimitString(finalReplacement, 100))
		return finalReplacement
	})
	resolveLogger.Debug("Handled !rep placeholders")

	resolveLogger.Debug("Finished resolving regex placeholders")
	resolveLogger.Trace("Final resolved pattern: %q", utils.LimitString(pattern, 100))
	return pattern
}

// ToLua sets capture groups as Lua variables (v1, v2, etc. for numeric values and s1, s2, etc. for strings)
func toLua(L *lua.LState, data interface{}) error {
	toLuaLogger := regexLogger.WithPrefix("toLua")
	toLuaLogger.Debug("Setting capture groups as Lua variables")

	captureGroups, ok := data.([]*CaptureGroup)
	if !ok {
		toLuaLogger.Error("Invalid data type for toLua. Expected []*CaptureGroup, got %T", data)
		return fmt.Errorf("expected []*CaptureGroup for captures, got %T", data)
	}
	toLuaLogger.Trace("Input capture groups: %v", captureGroups)

	groupindex := 0
	for _, capture := range captureGroups {
		groupLogger := toLuaLogger.WithField("captureGroup", capture.Name).WithField("value", utils.LimitString(capture.Value, 50))
		groupLogger.Debug("Processing capture group for Lua")

		if capture.Name == "" {
			// We don't want to change the name of the capture group
			// Even if it's empty
			tempName := fmt.Sprintf("%d", groupindex+1)
			groupindex++
			groupLogger.Debug("Unnamed capture group, assigning temporary name: %q", tempName)

			L.SetGlobal("s"+tempName, lua.LString(capture.Value))
			groupLogger.Trace("Set Lua global s%s = %q", tempName, capture.Value)

			val, err := strconv.ParseFloat(capture.Value, 64)
			if err == nil {
				L.SetGlobal("v"+tempName, lua.LNumber(val))
				groupLogger.Trace("Set Lua global v%s = %f", tempName, val)
			} else {
				groupLogger.Trace("Value %q is not numeric, skipping v%s assignment", capture.Value, tempName)
			}
		} else {
			val, err := strconv.ParseFloat(capture.Value, 64)
			if err == nil {
				L.SetGlobal(capture.Name, lua.LNumber(val))
				groupLogger.Trace("Set Lua global %s = %f (numeric)", capture.Name, val)
			} else {
				L.SetGlobal(capture.Name, lua.LString(capture.Value))
				groupLogger.Trace("Set Lua global %s = %q (string)", capture.Name, capture.Value)
			}
		}
	}

	toLuaLogger.Debug("Finished setting capture groups as Lua variables")
	return nil
}

// FromLua implements the Processor interface for RegexProcessor
func fromLua(L *lua.LState, captureGroups []*CaptureGroup) ([]*CaptureGroup, error) {
	fromLuaLogger := regexLogger.WithPrefix("fromLua")
	fromLuaLogger.Debug("Retrieving modifications from Lua for capture groups")
	fromLuaLogger.Trace("Initial capture groups: %v", captureGroups)

	captureIndex := 0
	for _, capture := range captureGroups {
		groupLogger := fromLuaLogger.WithField("originalCaptureName", capture.Name).WithField("originalValue", utils.LimitString(capture.Value, 50))
		groupLogger.Debug("Processing capture group to retrieve updated value")

		if capture.Name == "" {
			// This case means it was an unnamed capture group originally.
			// We need to reconstruct the original temporary name to fetch its updated value.
			// The name will be set to an integer if it was empty, then incremented.
			// So, we use the captureIndex to get the correct 'vX' and 'sX' variables.
			tempName := fmt.Sprintf("%d", captureIndex+1)
			groupLogger.Debug("Retrieving updated value for unnamed group (temp name: %q)", tempName)

			vVarName := fmt.Sprintf("v%s", tempName)
			sVarName := fmt.Sprintf("s%s", tempName)
			captureIndex++

			vLuaVal := L.GetGlobal(vVarName)
			sLuaVal := L.GetGlobal(sVarName)

			groupLogger.Trace("Lua values for unnamed group: v=%v, s=%v", vLuaVal, sLuaVal)

			if sLuaVal.Type() == lua.LTString {
				capture.Updated = sLuaVal.String()
				groupLogger.Trace("Updated value from s%s (string): %q", tempName, capture.Updated)
			}
			// Numbers have priority
			if vLuaVal.Type() == lua.LTNumber {
				capture.Updated = vLuaVal.String()
				groupLogger.Trace("Updated value from v%s (numeric): %q", tempName, capture.Updated)
			}
		} else {
			// Easy shit, directly use the named capture group
			updatedValue := L.GetGlobal(capture.Name)
			if updatedValue.Type() != lua.LTNil {
				capture.Updated = updatedValue.String()
				groupLogger.Trace("Updated value for named group %q: %q", capture.Name, capture.Updated)
			} else {
				groupLogger.Debug("Named capture group %q not found in Lua globals or is nil. Keeping original value.", capture.Name)
				capture.Updated = capture.Value // Keep original if not found or nil
			}
		}
		groupLogger.Debug("Finished processing capture group. Original: %q, Updated: %q", utils.LimitString(capture.Value, 50), utils.LimitString(capture.Updated, 50))
	}

	fromLuaLogger.Debug("Finished retrieving modifications from Lua")
	fromLuaLogger.Trace("Final updated capture groups: %v", captureGroups)
	return captureGroups, nil
}

// estimatePatternComplexity gives a rough estimate of regex pattern complexity
// This can help identify potentially problematic patterns
func estimatePatternComplexity(pattern string) int {
	estimateComplexityLogger := regexLogger.WithPrefix("estimatePatternComplexity").WithField("pattern", utils.LimitString(pattern, 100))
	estimateComplexityLogger.Debug("Estimating regex pattern complexity")
	complexity := len(pattern)

	// Add complexity for potentially expensive operations
	complexity += strings.Count(pattern, ".*") * 10 // Greedy wildcard
	complexity += strings.Count(pattern, ".*?") * 5 // Non-greedy wildcard
	complexity += strings.Count(pattern, "[^") * 3  // Negated character class
	complexity += strings.Count(pattern, "\\b") * 2 // Word boundary
	complexity += strings.Count(pattern, "(") * 2   // Capture groups
	complexity += strings.Count(pattern, "(?:") * 1 // Non-capture groups
	complexity += strings.Count(pattern, "\\1") * 3 // Backreferences
	complexity += strings.Count(pattern, "{") * 2   // Counted repetition

	estimateComplexityLogger.Debug("Estimated pattern complexity: %d", complexity)
	return complexity
}