Tser

A TypeScript virtual machine.

Switch Language: 中文.

Install and Usage

SupportMacOS 10.13+

$ brew tap tser-project/tser && brew install tser;

$ tser ./input.ts;

Why create Tser?

TypeScript (TS) is a greate invention, for letting us have a statically typed language development experience while reusing the JS ecosystem.
TS is essentially a pre-compiled language, executed using the JS virtual machine after compiled into JS. Due to its strong dependence on JS, it cannot get rid of some of the stubborn problems of JS, such as execution efficiency. TS itself is a statically typed language with certain data type declaration, but the type declaration is lost when escaping to JS; if we can directly execute the TS program instead of first escaping to JS and then executing, the data type declaration can Brings a great performance improvement to the program.

Let’s look at a set of performance comparison data, only comparing the performance of fib(42) in each virtual machine or language (It cannot be completely regarded as a performance evaluation standard. Test conditions: the same device in the same state, none of the compilation processes usded optimization).

Language	Virtual Machine	Time Cost (ms)
TypeScript	deno	4150
JavaScript	v8 / node	3859
TypeScript	Tser	2035
C++	–	2106

TS technology has developed rapidly, and more and more projects in the web industry have been developed and refactored using TS. When we thinking about the development of TS technology aside, where will it end? Will it stay on a pre-compiled language? When the TS ecosystem develops more and more sound, is it necessary to rely entirely on the JS ecosystem? Will there be a real TS virtual machine (Deno is not)? If the industry has a stable and high-performance TS virtual machine, is it a good thing for the TS ecosystem, and will it push TS to a new level?

TS may never be able to replace the JS ecology, but in some areas, TS can exist independently of the JS ecology; the TS virtual machine is the cornerstone of the independent TS ecology, allowing TS to have significantly better runtime performance than JS in these areas, and Bring real business benefits to these areas.

What can Tser do?

Tser has relatively low compilation performance and high runtime performance, which is more suitable for scenarios such as independent background services and serverless. When the grammar support is complete, and most of the existing background services written in TS can be run, then these services can get a great performance improvement.

Tser can do very little at present, because the syntax support is very imperfect, and can only support the execution of some simple scripts, for example, it may be able to support some simple cloud function scenarios.

Tser technology principle

The Tser’s front-end relies on the syntax parser generated by Antlr, and then generates and traverses the syntax tree; the back-end relies on LLVM construction, compiling TS code into LLVM IR, and using its JIT engine to immediately execute IR.

Similar products: AssemblyScript、StaticScript.

Tser syntax support

Refer to the __test file in the directory.

Tser’s grammar support will continue to improve.

Variable

Support var let const, var is the same as let.

Basic data type

Data Type	Bytes (in 64bit)	Supported
boolean	1	✔️
number	4	✔️ Same as int32
int32	4	✔️
int64	8	✔️
float	4	✔️
double	8	✔️
string	–	✔️ Not support operation

Operations

Operator	Supported
+ - * / %	✔️
++ --	✔️
+= -= *= /= %=	✔️
< > <= >=	✔️
== === != !==	✔️ === and == are the same currently
&&, ‖	✔️
!	✔️
? :	✔️
()	✔️
.	✔️

Logic control statement

statement	Supported
if else	✔️
while do while	✔️
for	✔️
switch	✔️
continue	✔️
break	✔️

Function

Support most parts of function, support nested functions, closure and function-as-parameter are temporarily not supported.

Example reference: __test/basic/function

Class

Class supports inheritance, polymorphism, class combination, supports static properties and static methods, and does not support method overloading; Class inheritance and polymorphism are supported in a manner similar to virtual table, and polymorphism supports methods and attributes.

Example reference: __test/class/

Built-in objects

The current built-in object support is very low and is only used for testing.

Built-in objects	methods
console	debug log info warn error
Date	now

Not supported currently

module Built-in objects、Event Loop、GC etc.

Contribute to Tser

Tser is a huge project and it is really difficult to complete it by one person. Now Tser is still a baby. Hope it can play a role of attractor which can gather some people with real abilities for better development.

Environmental setting

1. Install cmake、make
2. Install or build LLVM
3. Build Antlr、antlr4 build runtime
   • Create symlinks or copy files to lib/antlr, then the directory’s structure is:

     —— lib
       |—— antlr
         |—— include
           |—— antlr4-runtime
         |—— lib
           |—— libantlr4-runtime.a
         |—— share
     —— src
     

4. Install Boost: $ brew install boost
5. Environment variable configuration:

   # Clang
   export CC=/usr/bin/clang
   export CXX=/usr/bin/clang++
   
   # LLVM (Your path of llvm)
   export PATH="/usr/local/Cellar/llvm/10.0.1/bin:$PATH"
   
   # Antlr
   alias antlr4='java -jar /usr/local/lib/antlr-4.8-complete.jar'
   alias grun='java org.antlr.v4.gui.TestRig'
   export CLASSPATH=".:/usr/local/lib/antlr-4.8-complete.jar:$CLASSPATH"
   

Custom build

# Build
$ make build-parser
$ make build

# Usage
$ ./dist/bin/tser __test/basic/7_function.ts

Other command

• make build-debug : Build debugging package (Recommends using Debug mode to build LLVM while debugging)
• make build : Build a production package
• make build-builtin : Build built-in objects
• make build-parser : Rebuild the parser
• make generate-parser : Regenerate the parser’s code and grun tool
• make grun : Visually view the syntax tree generated by TS code

Donate to support

Donate to Tser, to support the author to continue to improve Tser.