_{Jamie Balfour's}Personal blog

After 5 years of development, the Zenith Parsing Engine is now being separated from the main ZPE application. This will make it easier to write compilers using it. As a result, I also intend to make separate JSON, XML and CSV compilers based on this. 

The reason for this is that I feel the Zenith Parsing Engine could be better utilised in other applications than including the whole package.

The very first version not to include the parser natively will be version 1.10.9, aka Lusty Lion.

ZPE is now available in native format!

Unlike previous operating-specific versions, these are not bundles but are actual binaries. They are compiled from the original JAR using GraalVM and native-image to improve performance. As a result, they are considerably larger files but also faster than the previous versions. A Linux version will also be available very soon. The JAR version will continue to be available and will remain as the primary version for those wanting the easiest of the installations.

As for Windows, I'm not sure when it will be available, since I only have one machine running Windows and I don't intend to turn it on to solely compile ZPE.

ZPE 1.10.8 is even faster than before! With new performance improvements and new features including TYPO v2.1 and typing within function declarations, ZPE 1.10.8 is set to be a big release. Data types have been added now to the stdLib to the functions that support it, but it became very apparent that multiple return types should be a bigger update than 1.10.7 was! 

How can we keep the performance high and yet add typing? Admittedly, every time a new feature is added, or a new check is required, performance takes a little dip. That is definitely true with anything. But performance can be made up elsewhere, and one of the areas where I have identified a performance gain is with the LAME evaluator and so I will continue to look into improving the LAME evaluator to make up for this.

Even without these improvements, or compiler optimisations, ZPE's standard library compiles and executes exceptionally fast, taking just 80ms to compile and run - that's very fast! 

Future plans for 1.10.8 also include auto prepending compiled files to make execution even faster. I will publish the next version of ZPE, which although will not be the finished version of ZPE 1.10.8, will allow you to test it out and I'll publish that today.

The ZPE/YASS compiler has featured support for typing of parameters for a long time, except it doesn't do anything. One could write something like the example below:

function test(number $s) {
  print($s)
}

test("Hello world")

And yet nothing will happen when it is given a string value instead of a number value. This is because, although it compiles correctly and is a well-formed ZPE expression, ZPE's runtime doesn't actually know what to do with it. 

Looking back, the way in which ZPE's function calls are made by placing parameters in an ordered set has barely changed and has stayed about the same since ZPE 1.3 when functions were properly introduced. 

ZPE also doesn't support return types. But it will very soon and the syntax has been decided (and included) already:

function test(number $s) : number {
  return $s + 5
}

Since the original method of generating parameters has been in ZPE and barely changed since 2015, it will be a rather big job to update this but it'll be worth it in the continuing move to making ZPE and YASS strongly typed. However, from ZPE 1.10.8 onwards, it will be possible to write programs that use strongly typed parameters and provide return types, they just will not do anything.