With BoxLang 1.7 released, one of the things we worked on was the performance of our parsing and bytecode generation. We’ve been improving this little-by-little for many releases, but now we actually have performance benchmarks comparing Adobe CF and Lucee Server.
We are pleased to announce that BoxLang’s parsing and compilation is FASTER than both Adobe CF 2025 and Lucee 6! (Lucee 7 wasn’t out when we finalized these tests)
Here are the comparisons. The ms represents the amount of time that engine spent parsing and compiling the test code. A lower value is better.
How did we measure? I created a simple test harness, which is published on GitHub for you to see and run yourself. Your numbers may vary depending on the speed and configuration of your PC, but the numbers here were captured on my Windows 11 PC with an SSD, 24 core i9 CPU and 64 Gigs of RAM. I ran all 3 engines on Java 21.
Here is the repo: GitHub - bdw429s/cf-compile-performance-testbed: My harness for testing the parse/compile performance of all 3 CF engines (Lucee, Adobe CF, BoxLang)
For a sizeable, but simple test, I parsed and compiled the entire ColdBox framework on all 3 engines. (Around 200 files) Prior to each test, I cleared the classes on disk and restarted the CF engine to get a cold start run of the parser with nothing cached or warmed up. I ran each test 3 times and averaged the results (with all the class clearing and restart between each test)
It’s worth noting that while BoxLang and Lucee have a way to JUST parse and compile a CFC or CFM, Adobe CF does not. Therefore, in my head-to-head tests with Lucee, I have both engines just parsing and generating bytecode. In my head-to-head tests with Adobe CF, I have both engines actually instantiating/including each file as this is the only way I know to get Adobe to parse and compile from within a CF-based harness. This means that the BoxLang numbers are different between each comparison, but each individual test should be as apples-to-apples as I can get it.
It’s also worth noting that ONLY Adobe and BoxLang support NOT writing class files to disk. In my tests, there was no noticeable performance difference when not writing class files, so I didn’t bother with that. All tests wrote class files to disk.
Here are the numbers for the graphs above:
BoxLang is 14 times faster than Adobe 2025 in our test
- BoxLang -
5,310 ms - Adobe -
76,601 ms
BoxLang is 2.6 times faster than Lucee 6.2.2 in our test
- BoxLang -
2,230 ms - Lucee -
5,865 ms
There are a lot of aspects that play into the performance of your application. This test obviously only focuses on that first-hit performance of compiling your code. We’ll do numbers on runtime performance soon, but this is just our first installment. We intend to make BoxLang as fast as possible in every area we can. We have an edge over Lucee, and we’re really blowing the doors off of Adobe CF in this area.
Related reading: BoxLang’s QoQ Is Here, And It’s 5x Faster Than Lucee, 17x Faster Than Adobe!