smaller jar with fewer dependencies?

[goodale]

I noticed the jar file is over 2MB. This seems very large for a parser. So not only increases war file size but also presumably adds additional vectors for security issues.

Is it possible to build a slim version that can be used in place of the older uadetector or uap-java, both of which are much smaller?

[nielsbasjes]

I have tried to minimise the dependencies and I have stripped as many of the unneeded classes as possible. So this os as small as it gets.

After startup this system uses a few hundred MiB of RAM, also a lot more than the others. And all of this makes the real goal possible: Best possible result in a time as short as possible.

So my main question is why is a 2MiB library a problem for you?

[nielsbasjes]

I pulled the jar apart and did some stats on the compressed content for version 7.22.0.

Summary: There is no way to reduce the size any further.

Group	Compressed bytes	Percentage of compressed file	Purpose
META-INF/	4614	0.2%	Needed in any jar
nl/basjes/parse/useragent/	402560	19.5%	The analyzer engine
nl/basjes/parse/useragent/parser/	224837	10.9%	The generated useragent parser
nl/basjes/parse/useragent/utils/publicsuffixlist/	7802	0.4%	Handling the domain suffixes
nl/basjes/parse/useragent/utils/springframework/	103203	5.0%	Dynamically finding the available yaml files
nl/basjes/shaded/org/antlr/	256797	12.4%	Runtime for the generated parser
nl/basjes/shaded/org/yaml/snakeyaml	273181	13.2%	Loading the yaml files.
UserAgents/	795374	38.5%	The yaml files with all the pattern rules.
Total Result	2068368	100.0%

[goodale]

thank you. I see a large part of the bundle is the yaml file, which makes sense. With some enterprise apps having upwards of hundreds of dependencies it is just prudent to use libs that are as small as possible, both to minimize start-up time but also publish time, etc. This is
a library that should be fine to filter from TLD scanning so that is one area to improve scan time. Thanks again.