mirror of
https://github.com/paperless-ngx/paperless-ngx.git
synced 2025-09-26 01:12:43 -05:00
Compare commits
1 Commits
feature-ex
...
release-wo
Author | SHA1 | Date | |
---|---|---|---|
![]() |
3cfb0f5856 |
2
.github/workflows/pr-bot.yml
vendored
2
.github/workflows/pr-bot.yml
vendored
@@ -37,7 +37,7 @@ jobs:
|
||||
labels.push('bug');
|
||||
} else if (/^feature/i.test(title)) {
|
||||
labels.push('enhancement');
|
||||
} else if (!/^(dependabot)/i.test(title) && !/^(chore)/i.test(title)) {
|
||||
} else if (!/^(dependabot)/i.test(title) && /^(chore)/i.test(title)) {
|
||||
labels.push('enhancement'); // Default fallback
|
||||
}
|
||||
|
||||
|
135
.github/workflows/release.yml
vendored
Normal file
135
.github/workflows/release.yml
vendored
Normal file
@@ -0,0 +1,135 @@
|
||||
name: Paperless-ngx Release
|
||||
on:
|
||||
workflow_dispatch:
|
||||
inputs:
|
||||
version:
|
||||
description: "Release version (e.g., 2.18.3)"
|
||||
required: true
|
||||
type: string
|
||||
permissions:
|
||||
contents: write
|
||||
actions: read
|
||||
concurrency:
|
||||
group: release-main
|
||||
cancel-in-progress: false
|
||||
jobs:
|
||||
release:
|
||||
runs-on: ubuntu-24.04
|
||||
steps:
|
||||
- name: Checkout (full)
|
||||
uses: actions/checkout@v4
|
||||
with:
|
||||
fetch-depth: 0
|
||||
- name: Configure git
|
||||
run: |
|
||||
git config user.name "${{ github.actor }}"
|
||||
git config user.email "${{ github.actor }}@users.noreply.github.com"
|
||||
- name: Sanitize & validate input
|
||||
id: ver
|
||||
shell: bash
|
||||
run: |
|
||||
RAW="${{ github.event.inputs.version }}"
|
||||
# trim spaces + strip leading 'v' if present
|
||||
RAW="${RAW//[[:space:]]/}"
|
||||
RAW="${RAW#v}"
|
||||
|
||||
# basic semver X.Y.Z
|
||||
if [[ ! "$RAW" =~ ^[0-9]+\.[0-9]+\.[0-9]+$ ]]; then
|
||||
echo "❌ Invalid version: '$RAW' (expected X.Y.Z or vX.Y.Z)"; exit 1
|
||||
fi
|
||||
|
||||
MAJOR="${RAW%%.*}"
|
||||
REST="${RAW#*.}"
|
||||
MINOR="${REST%%.*}"
|
||||
PATCH="${REST#*.}"
|
||||
|
||||
echo "version=$RAW" >> "$GITHUB_OUTPUT"
|
||||
echo "major=$MAJOR" >> "$GITHUB_OUTPUT"
|
||||
echo "minor=$MINOR" >> "$GITHUB_OUTPUT"
|
||||
echo "patch=$PATCH" >> "$GITHUB_OUTPUT"
|
||||
|
||||
echo "✅ Using version $RAW"
|
||||
- name: Ensure tag does not already exist
|
||||
run: |
|
||||
git fetch --tags
|
||||
if git rev-parse "v${{ steps.ver.outputs.version }}" >/dev/null 2>&1; then
|
||||
echo "❌ Tag v${{ steps.ver.outputs.version }} already exists"; exit 1
|
||||
fi
|
||||
- name: Update local branches
|
||||
run: |
|
||||
git fetch origin main dev
|
||||
- name: Fast-forward main to dev (no merge commits)
|
||||
run: |
|
||||
# Reset local main to remote, then try fast-forward to dev.
|
||||
git checkout main
|
||||
git reset --hard origin/main
|
||||
# --ff-only ensures the workflow fails if the branches diverged.
|
||||
git merge --ff-only origin/dev
|
||||
echo "✅ main fast-forwarded to dev at $(git rev-parse --short HEAD)"
|
||||
- name: Bump versions in files
|
||||
shell: bash
|
||||
run: |
|
||||
VER="${{ steps.ver.outputs.version }}"
|
||||
MAJ="${{ steps.ver.outputs.major }}"
|
||||
MIN="${{ steps.ver.outputs.minor }}"
|
||||
PAT="${{ steps.ver.outputs.patch }}"
|
||||
|
||||
# 1) pyproject.toml: [project] version = "X.Y.Z"
|
||||
sed -i -E 's/^version = "[0-9]+\.[0-9]+\.[0-9]+"/version = "'"$VER"'"/' pyproject.toml
|
||||
|
||||
# 2) src-ui/package.json: "version": "X.Y.Z"
|
||||
# Use jq if available; otherwise sed fallback.
|
||||
if command -v jq >/dev/null 2>&1; then
|
||||
tmp=$(mktemp)
|
||||
jq --arg v "$VER" '.version=$v' src-ui/package.json > "$tmp" && mv "$tmp" src-ui/package.json
|
||||
else
|
||||
sed -i -E 's/"version": "[0-9]+\.[0-9]+\.[0-9]+"/"version": "'"$VER"'"/' src-ui/package.json
|
||||
fi
|
||||
|
||||
# 3) src-ui/src/environments/environment.prod.ts: version: 'X.Y.Z'
|
||||
sed -i -E "s/version: '[0-9]+\.[0-9]+\.[0-9]+'/version: '$VER'/" src-ui/src/environments/environment.prod.ts
|
||||
|
||||
# 4) src/paperless/version.py: __version__ = (X, Y, Z)
|
||||
sed -i -E "s/__version__:\s*Final\[tuple\[int,\s*int,\s*int\]\]\s*=\s*\([0-9]+,\s*[0-9]+,\s*[0-9]+\)/__version__: Final[tuple[int, int, int]] = ($MAJ, $MIN, $PAT)/" src/paperless/version.py
|
||||
|
||||
# 5) uv.lock: in the [[package]] name = "paperless-ngx" block, set version = "X.Y.Z"
|
||||
# This awk edits only the block for paperless-ngx.
|
||||
awk -v ver="$VER" '
|
||||
BEGIN{inpkg=0}
|
||||
/^\[\[package\]\]/{inpkg=0}
|
||||
/^\[\[package\]\]/{print; next}
|
||||
{print > "/dev/stdout"}
|
||||
' uv.lock >/dev/null 2>&1 # noop to ensure awk exists
|
||||
|
||||
# More robust in-place edit with awk:
|
||||
awk -v ver="$VER" '
|
||||
BEGIN{inpkg=0}
|
||||
/^\[\[package\]\]/{inpkg=0; print; next}
|
||||
/^name = "paperless-ngx"/{inpkg=1; print; next}
|
||||
inpkg && /^version = "/{
|
||||
sub(/version = "[0-9]+\.[0-9]+\.[0-9]+"/, "version = \"" ver "\"")
|
||||
print; next
|
||||
}
|
||||
{print}
|
||||
' uv.lock > uv.lock.new && mv uv.lock.new uv.lock
|
||||
|
||||
echo "✅ Files updated to $VER"
|
||||
- name: Commit bump (if changes)
|
||||
run: |
|
||||
if git diff --quiet; then
|
||||
echo "ℹ️ No changes to commit (versions may already match)";
|
||||
else
|
||||
git add pyproject.toml src-ui/package.json src-ui/src/environments/environment.prod.ts src/paperless/version.py uv.lock
|
||||
git commit -m "Bump version to ${{ steps.ver.outputs.version }}"
|
||||
fi
|
||||
- name: Push main
|
||||
run: |
|
||||
# Push branch (even if no commit, ensures remote main == local)
|
||||
git push origin HEAD:main
|
||||
- name: Create and push tag
|
||||
run: |
|
||||
VER="${{ steps.ver.outputs.version }}"
|
||||
git tag -a "v${VER}" -m "Release v${VER}"
|
||||
git push origin "v${VER}"
|
||||
- name: Done
|
||||
run: echo "🎉 Release v${{ steps.ver.outputs.version }} created and pushed."
|
@@ -56,12 +56,12 @@
|
||||
"@playwright/test": "^1.55.0",
|
||||
"@types/jest": "^30.0.0",
|
||||
"@types/node": "^24.3.0",
|
||||
"@typescript-eslint/eslint-plugin": "^8.41.0",
|
||||
"@typescript-eslint/parser": "^8.41.0",
|
||||
"@typescript-eslint/utils": "^8.41.0",
|
||||
"eslint": "^9.34.0",
|
||||
"jest": "30.1.3",
|
||||
"jest-environment-jsdom": "^30.1.2",
|
||||
"@typescript-eslint/eslint-plugin": "^8.38.0",
|
||||
"@typescript-eslint/parser": "^8.38.0",
|
||||
"@typescript-eslint/utils": "^8.38.0",
|
||||
"eslint": "^9.32.0",
|
||||
"jest": "30.0.5",
|
||||
"jest-environment-jsdom": "^30.0.5",
|
||||
"jest-junit": "^16.0.0",
|
||||
"jest-preset-angular": "^15.0.0",
|
||||
"jest-websocket-mock": "^2.5.0",
|
||||
|
787
src-ui/pnpm-lock.yaml
generated
787
src-ui/pnpm-lock.yaml
generated
File diff suppressed because it is too large
Load Diff
@@ -2764,7 +2764,7 @@ class SystemStatusView(PassUserMixin):
|
||||
install_type = "docker"
|
||||
|
||||
db_conn = connections["default"]
|
||||
db_url = str(db_conn.settings_dict["NAME"])
|
||||
db_url = db_conn.settings_dict["NAME"]
|
||||
db_error = None
|
||||
|
||||
try:
|
||||
|
@@ -425,7 +425,7 @@ WHITENOISE_STATIC_PREFIX = "/static/"
|
||||
if machine().lower() == "aarch64": # pragma: no cover
|
||||
_static_backend = "django.contrib.staticfiles.storage.StaticFilesStorage"
|
||||
else:
|
||||
_static_backend = "paperless.staticfiles.DeduplicatedCompressedStaticFilesStorage"
|
||||
_static_backend = "whitenoise.storage.CompressedStaticFilesStorage"
|
||||
|
||||
STORAGES = {
|
||||
"staticfiles": {
|
||||
|
@@ -1,385 +0,0 @@
|
||||
import gzip
|
||||
import hashlib
|
||||
import logging
|
||||
import os
|
||||
import shutil
|
||||
import threading
|
||||
import time
|
||||
from collections import defaultdict
|
||||
from concurrent.futures import ThreadPoolExecutor
|
||||
from concurrent.futures import as_completed
|
||||
from dataclasses import dataclass
|
||||
from pathlib import Path
|
||||
|
||||
import brotli
|
||||
import humanize
|
||||
from django.contrib.staticfiles.storage import StaticFilesStorage
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
@dataclass(slots=True)
|
||||
class FileInfo:
|
||||
file_path_str: str
|
||||
file_path_path: Path
|
||||
checksum: str
|
||||
original_size: int
|
||||
gzip_size: int | None = None
|
||||
brotli_size: int | None = None
|
||||
|
||||
|
||||
class DeduplicatedCompressedStaticFilesStorage(StaticFilesStorage):
|
||||
# File extensions that should be compressed
|
||||
COMPRESSIBLE_EXTENSIONS = {
|
||||
".css",
|
||||
".js",
|
||||
".html",
|
||||
".htm",
|
||||
".xml",
|
||||
".json",
|
||||
".txt",
|
||||
".svg",
|
||||
".md",
|
||||
".rst",
|
||||
".csv",
|
||||
".tsv",
|
||||
".yaml",
|
||||
".yml",
|
||||
".map",
|
||||
}
|
||||
|
||||
# Minimum file size to compress (bytes)
|
||||
MIN_COMPRESS_SIZE = 1024 # 1KB
|
||||
|
||||
# Maximum number of threads for parallel processing
|
||||
MAX_WORKERS = min(32, (os.cpu_count() or 1) + 4)
|
||||
|
||||
# Chunk size for file reading
|
||||
CHUNK_SIZE = 64 * 1024 # 64KB
|
||||
|
||||
def __init__(self, *args, **kwargs):
|
||||
super().__init__(*args, **kwargs)
|
||||
# --- MODIFIED: Added path_to_file_info for easy lookup ---
|
||||
self.hash_to_files: dict[str, list[FileInfo]] = defaultdict(list)
|
||||
self.path_to_file_info: dict[str, FileInfo] = {}
|
||||
self.linked_files: set[Path] = set()
|
||||
self.compression_stats = {
|
||||
"brotli": 0,
|
||||
"gzip": 0,
|
||||
"skipped_linked": 0,
|
||||
"skipped_other": 0,
|
||||
"errors": 0,
|
||||
}
|
||||
self._lock = threading.Lock()
|
||||
|
||||
def post_process(self, paths: list[str], **options):
|
||||
"""
|
||||
Post-process collected files: deduplicate first, then compress.
|
||||
Django 5.2 compatible with proper options handling.
|
||||
"""
|
||||
start_time = time.time()
|
||||
|
||||
# Step 1: Build hash map for deduplication (parallel)
|
||||
self._build_file_hash_map_parallel(paths)
|
||||
|
||||
# Step 2: Create hard links for duplicate files
|
||||
self._create_hard_links()
|
||||
|
||||
# Step 3: Compress files (parallel, skip linked duplicates)
|
||||
self._compress_files_parallel(paths)
|
||||
|
||||
# Step 4: Provide user a summary of the compression
|
||||
self._log_compression_summary()
|
||||
|
||||
processing_time = time.time() - start_time
|
||||
logger.info(f"Post-processing complete in {processing_time:.2f}s.")
|
||||
|
||||
# Return list of processed files
|
||||
processed_files = []
|
||||
for path in paths:
|
||||
processed_files.append((path, path, True))
|
||||
# Add compressed variants
|
||||
file_path = self.path(path)
|
||||
if Path(file_path + ".br").exists():
|
||||
processed_files.append((path + ".br", path + ".br", True))
|
||||
if Path(file_path + ".gz").exists():
|
||||
processed_files.append((path + ".gz", path + ".gz", True))
|
||||
|
||||
return processed_files
|
||||
|
||||
def _build_file_hash_map_parallel(self, file_paths: list[str]):
|
||||
"""Build a map of file hashes using parallel processing."""
|
||||
logger.info(
|
||||
f"Hashing {len(file_paths)} files with {self.MAX_WORKERS} workers...",
|
||||
)
|
||||
|
||||
def hash_file(path: str):
|
||||
"""Hash a single file."""
|
||||
try:
|
||||
file_path = Path(self.path(path))
|
||||
if not file_path.is_file():
|
||||
return None, None, None
|
||||
|
||||
file_hash = self._get_file_hash_fast(file_path)
|
||||
file_size = file_path.stat().st_size
|
||||
return path, file_hash, file_size
|
||||
except Exception as e:
|
||||
logger.warning(f"Error hashing file {path}: {e}")
|
||||
return path, None, None
|
||||
|
||||
with ThreadPoolExecutor(max_workers=self.MAX_WORKERS) as executor:
|
||||
future_to_path = {
|
||||
executor.submit(hash_file, path): path for path in file_paths
|
||||
}
|
||||
|
||||
for future in as_completed(future_to_path):
|
||||
path, file_hash, file_size = future.result()
|
||||
if path is not None and file_hash is not None and file_size is not None:
|
||||
with self._lock:
|
||||
file_info = FileInfo(
|
||||
file_path_str=path,
|
||||
file_path_path=Path(self.path(path)),
|
||||
checksum=file_hash,
|
||||
original_size=file_size,
|
||||
)
|
||||
self.hash_to_files[file_hash].append(file_info)
|
||||
self.path_to_file_info[path] = file_info
|
||||
|
||||
duplicates = sum(1 for files in self.hash_to_files.values() if len(files) > 1)
|
||||
logger.info(f"Found {duplicates} sets of duplicate files")
|
||||
|
||||
def _get_file_hash_fast(self, file_path: Path):
|
||||
"""Calculate SHA-256 hash of file content with optimized reading."""
|
||||
hash_sha256 = hashlib.sha256()
|
||||
try:
|
||||
with file_path.open("rb") as f:
|
||||
while chunk := f.read(self.CHUNK_SIZE):
|
||||
hash_sha256.update(chunk)
|
||||
except OSError as e:
|
||||
logger.warning(f"Could not read file {file_path}: {e}")
|
||||
raise
|
||||
return hash_sha256.hexdigest()
|
||||
|
||||
def _create_hard_links(self):
|
||||
"""Create hard links for duplicate files."""
|
||||
logger.info("Creating hard links for duplicate files...")
|
||||
|
||||
linked_count = 0
|
||||
for file_info_list in self.hash_to_files.values():
|
||||
if len(file_info_list) <= 1:
|
||||
continue
|
||||
|
||||
# Sort by file size (desc) then path length (asc) to keep best original
|
||||
file_info_list.sort(key=lambda x: (-x.original_size, len(x.file_path_str)))
|
||||
original_file_info = file_info_list[0]
|
||||
duplicate_info = file_info_list[1:]
|
||||
|
||||
for duplicate_file_info in duplicate_info:
|
||||
try:
|
||||
# Remove duplicate file and create hard link
|
||||
if duplicate_file_info.file_path_path.exists():
|
||||
duplicate_file_info.file_path_path.unlink()
|
||||
|
||||
# Create hard link
|
||||
os.link(
|
||||
original_file_info.file_path_path,
|
||||
duplicate_file_info.file_path_path,
|
||||
)
|
||||
|
||||
with self._lock:
|
||||
self.linked_files.add(duplicate_file_info.file_path_path)
|
||||
|
||||
linked_count += 1
|
||||
|
||||
logger.info(
|
||||
f"Linked {duplicate_file_info.file_path_path} -> {original_file_info.file_path_path}",
|
||||
)
|
||||
|
||||
except OSError as e:
|
||||
logger.error(
|
||||
f"Hard link failed for {original_file_info.file_path_path}, copying instead: {e}",
|
||||
)
|
||||
# Fall back to copying if hard linking fails
|
||||
try:
|
||||
import shutil
|
||||
|
||||
shutil.copy2(
|
||||
original_file_info.file_path_path,
|
||||
original_file_info.file_path_path,
|
||||
)
|
||||
logger.error(
|
||||
f"Copied {original_file_info.file_path_path} (hard link failed)",
|
||||
)
|
||||
except Exception as copy_error:
|
||||
logger.error(
|
||||
f"Failed to copy {original_file_info.file_path_path}: {copy_error}",
|
||||
)
|
||||
|
||||
if linked_count > 0:
|
||||
logger.info(f"Created {linked_count} hard links")
|
||||
|
||||
def _compress_files_parallel(self, file_paths: list[str]):
|
||||
"""Compress files using parallel processing and update FileInfo objects."""
|
||||
# Identify files to compress, excluding hard links
|
||||
compressible_files = [
|
||||
self.path_to_file_info[path]
|
||||
for path in file_paths
|
||||
if self.path_to_file_info[path].file_path_path not in self.linked_files
|
||||
and self._should_compress_file(path)
|
||||
]
|
||||
|
||||
if not compressible_files:
|
||||
logger.info("No new files to compress")
|
||||
return
|
||||
|
||||
logger.info(
|
||||
f"Compressing {len(compressible_files)} files with {self.MAX_WORKERS} workers...",
|
||||
)
|
||||
|
||||
def compress_file(file_info: FileInfo):
|
||||
"""Compress a single file and update its FileInfo by side-effect."""
|
||||
brotli_size = None
|
||||
gzip_size = None
|
||||
error = None
|
||||
try:
|
||||
brotli_size = self._compress_file_brotli(str(file_info.file_path_path))
|
||||
gzip_size = self._compress_file_gzip(str(file_info.file_path_path))
|
||||
# Store the compressed sizes
|
||||
file_info.brotli_size = brotli_size
|
||||
file_info.gzip_size = gzip_size
|
||||
except Exception as e:
|
||||
error = str(e)
|
||||
logger.warning(f"Error compressing {file_info.file_path_str}: {e}")
|
||||
return {
|
||||
"brotli": brotli_size is not None,
|
||||
"gzip": gzip_size is not None,
|
||||
"error": error,
|
||||
}
|
||||
|
||||
with ThreadPoolExecutor(max_workers=self.MAX_WORKERS) as executor:
|
||||
future_to_info = {
|
||||
executor.submit(compress_file, info): info
|
||||
for info in compressible_files
|
||||
}
|
||||
|
||||
for future in as_completed(future_to_info):
|
||||
result = future.result()
|
||||
with self._lock:
|
||||
if result["brotli"]:
|
||||
self.compression_stats["brotli"] += 1
|
||||
if result["gzip"]:
|
||||
self.compression_stats["gzip"] += 1
|
||||
if result["error"]:
|
||||
self.compression_stats["errors"] += 1
|
||||
if (
|
||||
not result["brotli"]
|
||||
and not result["gzip"]
|
||||
and not result["error"]
|
||||
):
|
||||
self.compression_stats["skipped_other"] += 1
|
||||
|
||||
self.compression_stats["skipped_linked"] = len(self.linked_files)
|
||||
logger.info(f"File count stats: {self.compression_stats}")
|
||||
|
||||
def _should_compress_file(self, path: str):
|
||||
"""Determine if a file should be compressed."""
|
||||
file_ext = Path(path).suffix.lower()
|
||||
if file_ext not in self.COMPRESSIBLE_EXTENSIONS:
|
||||
return False
|
||||
try:
|
||||
if Path(self.path(path)).stat().st_size < self.MIN_COMPRESS_SIZE:
|
||||
return False
|
||||
except OSError:
|
||||
return False
|
||||
return True
|
||||
|
||||
def _compress_file_brotli(self, file_path: str) -> int | None:
|
||||
"""Compress file using Brotli, returns compressed size or None."""
|
||||
brotli_path = Path(file_path + ".br")
|
||||
try:
|
||||
with Path(file_path).open("rb") as f_in:
|
||||
original_data = f_in.read()
|
||||
compressed_data = brotli.compress(
|
||||
original_data,
|
||||
quality=10,
|
||||
lgwin=22, # Window size
|
||||
lgblock=0, # Auto block size
|
||||
)
|
||||
if len(compressed_data) < len(original_data) * 0.95:
|
||||
with brotli_path.open("wb") as f_out:
|
||||
f_out.write(compressed_data)
|
||||
return len(compressed_data)
|
||||
return None
|
||||
except Exception as e:
|
||||
logger.warning(f"Brotli compression failed for {file_path}: {e}")
|
||||
return None
|
||||
|
||||
def _compress_file_gzip(self, file_path: str) -> int | None:
|
||||
"""Compress file using GZip, returns compressed size or None."""
|
||||
gzip_path = Path(file_path + ".gz")
|
||||
file_path_path = Path(file_path)
|
||||
try:
|
||||
original_size = file_path_path.stat().st_size
|
||||
with (
|
||||
file_path_path.open("rb") as f_in,
|
||||
gzip.open(
|
||||
gzip_path,
|
||||
"wb",
|
||||
compresslevel=7,
|
||||
) as f_out,
|
||||
):
|
||||
shutil.copyfileobj(f_in, f_out, length=self.CHUNK_SIZE)
|
||||
|
||||
compressed_size = gzip_path.stat().st_size
|
||||
if compressed_size < original_size * 0.95:
|
||||
return compressed_size
|
||||
else:
|
||||
gzip_path.unlink()
|
||||
return None
|
||||
except Exception as e:
|
||||
logger.warning(f"GZip compression failed for {file_path}: {e}")
|
||||
if gzip_path.exists():
|
||||
try:
|
||||
gzip_path.unlink()
|
||||
except OSError:
|
||||
pass
|
||||
return None
|
||||
|
||||
def _log_compression_summary(self):
|
||||
"""Calculates and logs the total size savings from compression."""
|
||||
total_original_size = 0
|
||||
total_brotli_size = 0
|
||||
total_gzip_size = 0
|
||||
|
||||
# Only consider the original files, not the duplicates, for size calculation
|
||||
unique_files = {
|
||||
file_list[0].checksum: file_list[0]
|
||||
for file_list in self.hash_to_files.values()
|
||||
}
|
||||
|
||||
for file_info in unique_files.values():
|
||||
if self._should_compress_file(file_info.file_path_str):
|
||||
total_original_size += file_info.original_size
|
||||
if file_info.brotli_size:
|
||||
total_brotli_size += file_info.brotli_size
|
||||
if file_info.gzip_size:
|
||||
total_gzip_size += file_info.gzip_size
|
||||
|
||||
def get_savings(original: int, compressed: int) -> str:
|
||||
if original == 0:
|
||||
return "0.00%"
|
||||
return f"{(1 - compressed / original) * 100:.2f}%"
|
||||
|
||||
logger.info(
|
||||
f"Total Original Size (compressible files): {humanize.naturalsize(total_original_size)}",
|
||||
)
|
||||
if total_brotli_size > 0:
|
||||
logger.info(
|
||||
f"Total Brotli Size: {humanize.naturalsize(total_brotli_size)} "
|
||||
f"(Savings: {get_savings(total_original_size, total_brotli_size)})",
|
||||
)
|
||||
if total_gzip_size > 0:
|
||||
logger.info(
|
||||
f"Total Gzip Size: {humanize.naturalsize(total_gzip_size)} "
|
||||
f"(Savings: {get_savings(total_original_size, total_gzip_size)})",
|
||||
)
|
Reference in New Issue
Block a user