refactor: ai cleanup

Author: iloveitaly

core/cleanse.go

@@ -1,40 +1,30 @@
 package core
 
-// This file contains logic related to post-generation cleansing / normalization
-// of build plans. Keeping this separate from `core.go` keeps provider selection
-// and config merge logic focused while allowing targeted tests for plan
-// mutation behavior.
-
 import (
 	"regexp"
 
-	"github.com/charmbracelet/log"
 	"github.com/railwayapp/railpack/core/logger"
 	"github.com/railwayapp/railpack/core/plan"
+	"github.com/railwayapp/railpack/core/providers/node"
 )
 
 // Regexes for matching commands that intentionally remove node_modules or perform
 // clean installs (which implicitly delete the directory) so we can avoid mounting
 // the node_modules cache in those steps.
 var (
-	npmCiCommandRegex      = regexp.MustCompile(`.*npm\s+ci\b.*`)
-	removeNodeModulesRegex = regexp.MustCompile(`(^|\s)(rm\s+-rf\s+|rimraf\s+)(\./)?node_modules(\s|;|&|$)`)
-)
+	// Matches "npm ci" with flexible whitespace, using word boundaries
+	npmCiCommandRegex = regexp.MustCompile(`(?i)\bnpm\s+ci\b`)
 
-const (
-	// NODE_MODULES_CACHE is the path inside the build container we treat as the
-	// node modules cache. Steps which delete node_modules should not mount this
-	// cache so the deletion stays local to the layer they're producing and does
-	// not wipe the shared cache directory.
-	NODE_MODULES_CACHE = "/app/node_modules/.cache"
+	// Matches common delete commands targeting node_modules
+	removeNodeModulesRegex = regexp.MustCompile(`(?i)\b(?:rm\s+-r[f]?|rmdir|rimraf)\s+(?:\S*\/)?node_modules\b`)
 )
 
 // willRemoveNodeModules determines if any command in the provided slice removes
 // the node_modules directory either directly (rm/rimraf) or indirectly (npm ci).
+// this is brittle & imperfect: https://github.com/railwayapp/railpack/pull/259
 func willRemoveNodeModules(commands []plan.Command) bool {
 	for _, cmd := range commands {
 		if execCmd, ok := cmd.(plan.ExecCommand); ok {
-			log.Debugf("Inspecting build command: %s", execCmd.Cmd)
 			if npmCiCommandRegex.MatchString(execCmd.Cmd) || removeNodeModulesRegex.MatchString(execCmd.Cmd) {
 				return true
 			}
@@ -51,7 +41,7 @@ func cleansePlanStructure(buildPlan *plan.BuildPlan, logger *logger.Logger) {
 	// let's get the cache key name that has a Directory of NODE_MODULES_CACHE
 	var nodeModulesCacheKey string
 	for cacheName, cacheDef := range buildPlan.Caches {
-		if cacheDef.Directory == NODE_MODULES_CACHE {
+		if cacheDef.Directory == node.NODE_MODULES_CACHE {
 			nodeModulesCacheKey = cacheName
 			break
 		}
@@ -74,6 +64,7 @@ func cleansePlanStructure(buildPlan *plan.BuildPlan, logger *logger.Logger) {
 			continue
 		}
 
+		// It's important that we do not result in an array with a zeroed string, which is why we are using this ugly loop
 		var newCaches []string
 		for _, name := range step.Caches {
 			if name != "" && name != nodeModulesCacheKey {