Hi, new release 0.5.1 is out for testing here.
Release notes:
- Complete rework of compile-time value representation.
- Command line arguments are now presented in command_line_arguments global.
- Custom executable startup (removed dependency on crt on macos and linux).
- Structure inheritance.
- Switch statement.
- No need to write semicolons after functions and typedefs in global scope.
- Structure members are now separated by semicolon.
- Enum variants are now separated by semicolon.
- Bugfixes: #65, #62, #63, #57, #64, #39, #61, #68, #67
Release notes:
- Complete rework of compile-time value representation.
- Command line arguments are now presented in command_line_arguments global.
- Custom executable startup (removed dependency on crt on macos and linux).
- Structure inheritance.
- Switch statement.
- No need to write semicolons after functions and typedefs in global scope.
- Structure members are now separated by semicolon.
- Enum variants are now separated by semicolon.
- Bugfixes: #65, #62, #63, #57, #64, #39, #61, #68, #67
Edited by Martin Dorazil
on
Reason: Initial post
Hi, there is new BL version 0.6.0 out for testing. HERE
Release notes:
- Add implicit cast from array to slice. (Fixed compile time vs constant IR code emit).
- Fix invalid conversion of the sizeof result to Any.
- Fix local functions in nested scope.
- Fix nested function implicitly capturing members of upper scope.
- Fix segfault on global initializer nesting.
- Fix LLVM cast assertions on string array to slice conversion.
- Fix invalid sizeof operator result on array elem values and struct memers.
- Fix lambda functions inside nested scope.
- Add implicit conversion from pointers to bool.
- Add union data type.
- Add support of Windows terminal color output.
- Add testing formaters for printing float values.
- Add support of MSVC CodeView and PDB debug database generation via LLVM API.
Release notes:
- Add implicit cast from array to slice. (Fixed compile time vs constant IR code emit).
- Fix invalid conversion of the sizeof result to Any.
- Fix local functions in nested scope.
- Fix nested function implicitly capturing members of upper scope.
- Fix segfault on global initializer nesting.
- Fix LLVM cast assertions on string array to slice conversion.
- Fix invalid sizeof operator result on array elem values and struct memers.
- Fix lambda functions inside nested scope.
- Add implicit conversion from pointers to bool.
- Add union data type.
- Add support of Windows terminal color output.
- Add testing formaters for printing float values.
- Add support of MSVC CodeView and PDB debug database generation via LLVM API.
Edited by Martin Dorazil
on
Hi, new release 0.9.0 is out! here
Release notes:
- Use LLD as default linker on Windows and Linux (experimental on macOS).
- Update command line arguments to match unix common naming standarts.
- Improve build system api and remove some obsolete functions.
- Change build pipeline to call compile/compile_all explicitly.
- Fix switch related parser bugs.
- Fix unit mixing when using multithreaded build pipeline.
- Add support of shared library output.
- Add cmake build target Utils.
- Add hash into CodeLocation.
- Add namespace support.
- Add unnamed symbols. (underscore used as name is not inserted into scope)
- Add support of inline function group members.
- Use faster custom system library and SDK lookup on Windows.
- Compiler now reports unused private or local symbols.
- Cleanup modules.
Release notes:
- Use LLD as default linker on Windows and Linux (experimental on macOS).
- Update command line arguments to match unix common naming standarts.
- Improve build system api and remove some obsolete functions.
- Change build pipeline to call compile/compile_all explicitly.
- Fix switch related parser bugs.
- Fix unit mixing when using multithreaded build pipeline.
- Add support of shared library output.
- Add cmake build target Utils.
- Add hash into CodeLocation.
- Add namespace support.
- Add unnamed symbols. (underscore used as name is not inserted into scope)
- Add support of inline function group members.
- Use faster custom system library and SDK lookup on Windows.
- Compiler now reports unused private or local symbols.
- Cleanup modules.
When your language has enums but no OOP, adding table-driven features would allow:
* Removing repeated (if x equals this or that value) if statements (look up a flag from an enum's table instead)
* Reduce the need for switch cases (get a function pointer from the table and reuse it multiple times)
* Improve performance (because it's better to read one commonly accessed word from D-cache than reading lots of long machine instructions scattered over megabytes in I-cache and jumping around randomly)
* Reduce how many mistakes are made when code changes (because each new enum value must fill all table properties)
* Multi-table lookups with clean syntax according to the third normal form. ("weapon.type.ammoType.calibre" becomes "ammoTable[weaponTable[weapon.type].ammoType].calibre")
The error-prone hard-coded way (imagine having this repeated on hundreds of locations before adding a new vehicle type in the enum).
Even if you put this in a function for each property, it takes up a lot of space and is rarely actually used.
The table driven way.
You can then just add another row in your table when something new has to be handled.
Feel free to borrow any other features from my language.
http://uu.diva-portal.org/smash/get/diva2:752088/FULLTEXT01
* Removing repeated (if x equals this or that value) if statements (look up a flag from an enum's table instead)
* Reduce the need for switch cases (get a function pointer from the table and reuse it multiple times)
* Improve performance (because it's better to read one commonly accessed word from D-cache than reading lots of long machine instructions scattered over megabytes in I-cache and jumping around randomly)
* Reduce how many mistakes are made when code changes (because each new enum value must fill all table properties)
* Multi-table lookups with clean syntax according to the third normal form. ("weapon.type.ammoType.calibre" becomes "ammoTable[weaponTable[weapon.type].ammoType].calibre")
The error-prone hard-coded way (imagine having this repeated on hundreds of locations before adding a new vehicle type in the enum).
Even if you put this in a function for each property, it takes up a lot of space and is rarely actually used.
1 2 3 | if (vehicleType == car || vehicleType == mc || vehicleType == boat || ...) {
// For those who have wheels or a steering wheel (The programmer should at least know why certain vehicles are selected)
}
|
The table driven way.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 | // Each table creates either a packed structure array for each value or one planar array per property depending on access pattern
enum VehicleType {
key int numberOfWheels bool hasSteeringWheel ...
car 4 true
mc 2 false
boat 0 true
...
}
/*
Using an attribute lookup on an integer translates to an array lookup
so that "vehicleType.numberOfWheels" becomes
"VehicleType_table[vehicleType].numberOfWheels" or "VehicleType_numberOfWheels[vehicleType]"
*/
if (vehicleType.numberOfWheels > 0 || vehicleType.hasSteeringWheel) {
// For those who have wheels or a steering wheel
}
|
You can then just add another row in your table when something new has to be handled.
Feel free to borrow any other features from my language.
http://uu.diva-portal.org/smash/get/diva2:752088/FULLTEXT01
Edited by Dawoodoz
on