imgstat 2.0.4 → 2.0.5

package/lib/analyze.sh

@@ -5,88 +5,140 @@ source "$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)/utils.sh"
 cmd_analyze() {
   local dir="$1"
   echo "Analyzing codebase for remote image references in $dir..."
-  
-  # Scan common code files for URLs
-  # We extract http/https links specifically in src, href, url() contexts
-  local urls=()
-  
-  # Read all matching files, use grep to find URLs, and sort uniquely
-  # This uses a basic grep regex to find common image URL patterns.
-  mapfile -t urls < <(find "$dir" \
-    -type d \( -name "node_modules" -o -name ".git" -o -name "dist" -o -name "build" \) -prune -o \
-    -type f \( -name "*.html" -o -name "*.jsx" -o -name "*.tsx" -o -name "*.js" -o -name "*.ts" -o -name "*.vue" -o -name "*.css" -o -name "*.scss" \) \
+
+  # ── Step 1: Extract all http/https URLs from common code files ──────────────
+  local all_urls=()
+  mapfile -t all_urls < <(find "$dir" \
+    -type d \( -name "node_modules" -o -name ".git" -o -name "dist" -o -name "build" -o -name "vendor" -o -name ".next" -o -name "coverage" \) -prune -o \
+    -type f \( -name "*.html" -o -name "*.jsx" -o -name "*.tsx" -o -name "*.js" -o -name "*.ts" \
+               -o -name "*.vue" -o -name "*.css" -o -name "*.scss" -o -name "*.php" \
+               -o -name "*.py" -o -name "*.rb" -o -name "*.svelte" -o -name "*.astro" \) \
     -exec grep -oE "https?://[^\"')[:space:]]+" {} + 2>/dev/null | sort -u)
 
-  if [[ ${#urls[@]} -eq 0 ]]; then
+  if [[ ${#all_urls[@]} -eq 0 ]]; then
     echo "No remote URLs found in code files."
     return 0
   fi
-  
-  echo "Found ${#urls[@]} unique URL(s). Fetching dimensions..."
-  
-  # Prepare secure fetch directory
+
+  echo "Found ${#all_urls[@]} unique URL(s). Classifying..."
+  echo ""
+
+  # ── Step 2: Classify each URL via 3-tier pipeline ──────────────────────────
+  local image_extensions=("jpg" "jpeg" "png" "gif" "webp" "avif" "svg" "bmp" "tiff" "tif" "ico")
+
+  local queued_urls=()
+
+  for url in "${all_urls[@]}"; do
+    # Strip query/fragment to inspect the path cleanly
+    local path
+    path=$(echo "$url" | sed 's/[?#].*//')
+    local lower_path
+    lower_path=$(echo "$path" | tr '[:upper:]' '[:lower:]')
+
+    # ── Tier 1: Obvious non-image check (fast, no network) ──────────────────
+    if is_obvious_non_image "$url"; then
+      printf "  \033[2m⏭  Skip  (pattern match) : %s\033[0m\n" "$url"
+      continue
+    fi
+
+    # ── Tier 2: Image extension in path (fast, no network) ───────────────────
+    local ext
+    ext=$(echo "${lower_path##*.}" | sed 's/[^a-z]//g')
+    local is_image_ext=false
+    for img_ext in "${image_extensions[@]}"; do
+      if [[ "$ext" == "$img_ext" ]]; then
+        is_image_ext=true
+        break
+      fi
+    done
+
+    if [[ "$is_image_ext" == "true" ]]; then
+      printf "  \033[1;32m✓  Queue (extension)     : %s\033[0m\n" "$url"
+      queued_urls+=("$url")
+      continue
+    fi
+
+    # ── Tier 3: HTTP HEAD Content-Type check (network, no body download) ─────
+    printf "  \033[33m?  Check (HEAD request)  : %s\033[0m" "$url"
+    if check_content_type_is_image "$url"; then
+      printf "\r  \033[1;32m✓  Queue (content-type)  : %s\033[0m\n" "$url"
+      queued_urls+=("$url")
+    else
+      printf "\r  \033[2m⏭  Skip  (not image CT)  : %s\033[0m\n" "$url"
+    fi
+  done
+
+  echo ""
+
+  if [[ ${#queued_urls[@]} -eq 0 ]]; then
+    echo "No image URLs found after classification."
+    return 0
+  fi
+
+  echo "Fetching dimensions for ${#queued_urls[@]} image URL(s)..."
+  echo ""
+
+  # ── Step 3: Download & measure queued image URLs ───────────────────────────
   local TEMP_DIR
   TEMP_DIR=$(mktemp -d)
   trap 'rm -rf "$TEMP_DIR"' EXIT
-  
+
   # Output file setup
   local rules_dir="$dir/.agent/rules"
   local rules_file="$rules_dir/image_dimensions.md"
-  
   mkdir -p "$rules_dir"
-  
-  # Write header for the agent
+
+  # Write header
   cat << 'EOF' > "$rules_file"
 # Codebase Remote Images
 
 > [!NOTE]
-> This file is strictly auto-generated by \`imgstat\`.
+> This file is strictly auto-generated by `imgstat`.
 > It maps remote image URLs found in the codebase to their exact physical dimensions.
 > AI assistants should use this dictionary when asked about layout constraints, native sizes, or aspect ratios of referenced images.
 
 | Documented URL | Detected Size (W x H) |
 |---|---|
 EOF
-  
+
   local processed=0
-  
-  for url in "${urls[@]}"; do
-    # Skip obvious non-images if possible, but Cloudinary etc don't have extensions
-    # so we attempt a shallow fetch for all to be safe.
-    
-    # 1. Fetch direct URL
-    wget -q --content-disposition -P "$TEMP_DIR" "$url" || true
-    
+
+  for url in "${queued_urls[@]}"; do
+    # Download the image content only
+    wget -q --content-disposition --max-redirect=5 \
+      --user-agent="Mozilla/5.0" \
+      -P "$TEMP_DIR" "$url" 2>/dev/null || true
+
     local all_files=( "$TEMP_DIR"/* )
-    local found_images=()
-    
+    local found_image=""
+
     for file in "${all_files[@]}"; do
-      if [[ ! -f "$file" ]]; then continue; fi
+      [[ ! -f "$file" ]] && continue
       local mimetype
       mimetype=$(file -b --mime-type "$file" 2>/dev/null || true)
       if [[ "$mimetype" == image/* ]]; then
-        found_images+=("$file")
+        found_image="$file"
+        break
       fi
     done
-    
-    # Process only the first valid image found for the URL
-    if [[ ${#found_images[@]} -gt 0 ]]; then
-      local file="${found_images[0]}"
+
+    if [[ -n "$found_image" ]]; then
       local dim
-      dim=$(get_dimensions "$file" || true)
-      
+      dim=$(get_dimensions "$found_image" || true)
       if [[ -n "$dim" ]]; then
         local w="${dim% *}"
         local h="${dim#* }"
         echo "| \`$url\` | **${w}x${h}** |" >> "$rules_file"
+        printf "  \033[1;32m📐 %s → %sx%s\033[0m\n" "$url" "$w" "$h"
         processed=$((processed + 1))
       fi
     fi
-    
-    # Clean temp dir contents for the next URL to avoid collision
+
+    # Clean temp dir for next URL to avoid collisions
     rm -rf "${TEMP_DIR:?}"/*
   done
-  
+
+  echo ""
   echo "Analysis complete!"
   echo "Documented $processed valid image reference(s) into: $rules_file"
   echo "This file can now be read by autonomous coding agents."

package/lib/utils.sh

@@ -62,3 +62,84 @@ format_filename() {
   
   echo "$dirname/$new_name"
 }
+
+# Returns 0 (true) if the URL is obviously NOT an image.
+# Uses domain and path-extension heuristics — no network call needed.
+# Usage: is_obvious_non_image "https://..."
+is_obvious_non_image() {
+  local url="$1"
+
+  # Strip query string / fragment for path inspection
+  local path
+  path=$(echo "$url" | sed 's/[?#].*//')
+
+  # --- Known non-image domains ---
+  local non_image_domains=(
+    "fonts.googleapis.com"
+    "fonts.gstatic.com"
+    "googletagmanager.com"
+    "google-analytics.com"
+    "www.google-analytics.com"
+    "youtube.com"
+    "www.youtube.com"
+    "youtu.be"
+    "vimeo.com"
+    "www.vimeo.com"
+    "maps.google.com"
+    "maps.googleapis.com"
+    "plausible.io"
+    "cdn.rawgit.com"
+    "ajax.googleapis.com"
+    "cdnjs.cloudflare.com"
+  )
+  local domain
+  domain=$(echo "$url" | sed -E 's|https?://([^/]+).*|\1|')
+  for d in "${non_image_domains[@]}"; do
+    if [[ "$domain" == "$d" ]]; then
+      return 0
+    fi
+  done
+
+  # --- Non-image path extensions ---
+  local lower_path
+  lower_path=$(echo "$path" | tr '[:upper:]' '[:lower:]')
+  local non_image_exts=(".js" ".mjs" ".css" ".scss" ".json" ".xml" ".html"
+    ".woff" ".woff2" ".ttf" ".eot" ".otf"
+    ".pdf" ".zip" ".tar" ".gz"
+    ".mp4" ".mp3" ".ogg" ".webm" ".avi" ".mov"
+    ".txt" ".md" ".csv"
+  )
+  for ext in "${non_image_exts[@]}"; do
+    if [[ "$lower_path" == *"$ext" ]]; then
+      return 0
+    fi
+  done
+
+  # --- Non-image URL path fragments ---
+  local non_image_patterns=("/api/" "/feed" "/rss" "/sitemap" "/graphql" "/oauth" "/auth/" "/login" "/logout")
+  for pat in "${non_image_patterns[@]}"; do
+    if [[ "$lower_path" == *"$pat"* ]]; then
+      return 0
+    fi
+  done
+
+  return 1
+}
+
+# Returns 0 (true) if the URL's HTTP Content-Type header indicates an image.
+# Uses a lightweight HEAD request — no body downloaded.
+# Usage: check_content_type_is_image "https://..."
+check_content_type_is_image() {
+  local url="$1"
+  local content_type
+  content_type=$(curl -sI --max-time 8 --location \
+    -H "User-Agent: Mozilla/5.0" \
+    "$url" 2>/dev/null \
+    | grep -i '^content-type:' \
+    | tail -1 \
+    | tr -d '\r')
+  if [[ "$content_type" == *"image/"* ]]; then
+    return 0
+  fi
+  return 1
+}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "imgstat",
-  "version": "2.0.4",
+  "version": "2.0.5",
   "description": "Embeds image dimensions directly into filenames for natural AI context.",
   "scripts": {
     "test": "echo \"Error: no test specified\" && exit 1"