Smally

• How to Compress
  • JPEG
  • PNG
  • GIF
• How to Install
• How to Use
• APIs
• Showcase

Smally is a command line tool which can help you compress JPEG, PNG and GIF files losslessly, by invoking the famous 3rd party tools such as jpegtran, optipng and gifsicle, in batch and parallel mode, inplace and keep mtime unchanged. It is written in Python, combined with SQLite, and heavily relied on shell.

Actually, the code structure and logic of smally might be a bit strange. It repeated invoke itself throught a shell command which is triggered by a subprocess initiated by itself as well. I just want to keep everything as simple and straight forward as possible. In some scenarios it indeed does a little bit more unnecessary work, but t's definitely also very funny! :)

How to Compress

JPEG

1. Use jpegtran to remove all metadata, create a baseline version and a progressive version.
2. Compare among the original file, baseline and progressive files, choose the smallest one in size.
3. Whenever possible, choose progressive version.

PNG

Call optipng to compress PNG, in the most crazy -o7 -zm1-9 level. (Caveat: time-consuming, but better compression ratio)

GIF

Call gifsicle to compress GIF, by using -O3 --color 256.

How to Install

You need to install all 3rd party tools and smally itself as well.

# install tools on Fedora
$ sudo dnf install libjpeg-turbo-utils optipng gifsicle
# install tools on Ubuntu
$ sudo apt install libjpeg-turbo-progs optipng gifsicle
# install smally
$ pip install smally

How to Use

# command line help
$ python -m smally -h
# compress a single file, if option is not presented,
# smally will use file command to get file type info.
$ python -m smally [-j|-p|-g] <pathname>
# to compress a directory
$ python -m smally -r -P<N> <pathname>
# to compress all png and gif file in a directory
$ python -m smally -r -P<N> -p -g <pathname>

-r, recursive, it's a command line convention and normally you should use it when deal with a directory.

-P<N>, number of parallel processes, if missing, the logical cpu number would be used by defalut. If the specific number is larger than the number of logical cpu, smally choose the smaller one, since compression is cpu-intensive job, it makes no sense that the process number is bigger than the number of cpu.

# delete database record recursively
$ python -m smally -r -d <pathname>
# delete database record recursively only for jpeg picture
$ python -m smally -P4 -r -d -j <pathname>
# delete all temporary files generated by smally recursively
$ python -m smally -r -c <pathname>

-d, delete database record for the specific pathname when you want to redo the compress job, since the purpose of database records is to prevent smally from redoing the same job.

-c, remove all temporary files generated by smally, which includes database file (SQLite) and database lock file (I use a file lock to realize mutex for multiple processes). If you want to redo everything, or store all your pictures to a place, use it.

APIs

There are 4 APIs offered by smally:

# import
from smally import (jpegtran,
                    optipng,
                    gifsicle,
                    is_jpeg_progressive)

# signature
def jpegtran(pathname: str) -> tuple[int,int]: ...
def optipng(pathname: str) -> tuple[int,int]: ...
def gifsicle(pathname: str) -> tuple[int,int]: ...
def is_jpeg_progressive(pathname: str) -> bool: ...

The first int in returned tuple is the saved byte number. It could be zero, which means no save, or negetive, which mean how many bytes have been saved. The second int is the original file size.

Showcase