Hi Boriel,

I wrote the following little game (which is supposed to become a "Jewel Quest" clone one day :-):

Changes from Version 1.1.2 to 1.1.4

• + The peephole optimizer has been enabled and seems to be working
  reasonabily well (use -O3 for higher optimization).
  
• ! When a DIV BY ZERO error occurs with floating point numbers the program
  crashes and resets the computer. This behaviour has been fixed
  and now returns 0 value and sets the error code 6 (Numbert Too big)
  in the ERR_NR system variable.
  
• * Refactorization of the assembler and compiler so they now
  shared the OPTIONS in a better way (still to be finished). This
  makes easier to program future compiler options. Now also --debug
  flag is additive, showing more verbosity the more it is used.
  
• + Memory optimization: The PRINT routine (which is about 1K) is not
  included if not USED.
  
• ! Fixed a bug in negative constant integer typecasting (Thanks to LCD
  at WOS for reporting it! ;-)). It was causing decremental FOR..NEXT
  to fail.
  
• ! Scientific notation numbers (e.g. 2e10) were not correctly parsed.
  Fixed. Thanks again to LCD. ;-)
  
• + Added TAB compatibility for the PRINT command (both as a command
  and as a CHR$ control character).
  
• * PRINT code optimized for size, maintaining speed.

• Some little more peephole optimizations.
  
• Implementation of SAVE/LOAD/VERIFY with DATA/CODE/SCREENS (And maybe even the BASIC zone)
  
• 8 bits AND/OR/NOT/XOR instructions
  
• READ/DATA/RESTORE (mostly for compatibility)
  
• CLEAR for filling memory zones.
  
• USR n and USR n$ function (mostly for compatibilty)

This version fixed some issues and introduces interesting improvements:

Changes from Version 1.1.0 to 1.1.1

• !Fixed a bug in CONTINUE DO which was not being correctly compiled
  
• +PRINT routines were included even when neither PRINT nor
  drawing primitives were used. Optimized.[/*]
  
• ! Fixed a lot of syntax error checkings with array operations.
  
• ! Fixed array dimension checking
  
• + Expanded syntax: Direct array assignation a = b (being a and b
  arrays of thes same type an size)[/*]
  
• ! Fixes an error exception on syntax error for array subscripting.
  
• * Changed alloc functions to match the FreeBasic names.
  
• * Using a wrong sigil in array declaration is now forbidden.
  
• * Better sigils (suffixes) types managements at DIM declarations.
  
• * Lot of internal source code refactoring
  
• + DIM r AT @a(k0, k1, ...) is allowed (k0, k1, ... constants)
  
• ! Fixed a bug for local variables and parameters when the offset is
  very large (> 128)
  
• + Enabled the --sinclair flag for automatic sinclair libraries inclusion
  
• + Added SetAttr routine whichs changes the attribute of
  screen coordinate (I, J) with the given color attr
  value.
  
• ! Fixed a buggy modu16 modi16 implementation that was not compiling
  correctly
  
• * Output asm code is now slightly optimized (for speed an memory)
  
• ! Fixed a bug in intenger parameters (16 and 32 bits)
  
• ! Fixed a compiler crash when using arrays of Fixed Point numbers

Hi Boriel, I just trying to write some stuff in ZX Basic compiler 1.1.0, but I ran into some problems:

This version fixed some issues and introduces interesting improvements:

Changes from Version 1.0.9 to 1.1.0

• SCREEN$ coordinated were swapped. Fixed.
  
• DIM .. AT was not correctly working with local vars nor params. Fixed.
  
• Added BOLD "attribute".
  Code:

  PRINT BOLD 1; "Hello"
• Added ITALIC "attribute".
  Code:

  PRINT ITALIC 1; "Hello"
• Added malloc, free and realloc functions to work with the heap
  
• Some code rearrangement
  
• The IFDEF directive was not working in the preprocessor. Fixed.
  

• Fixed a bug which could crash the program if no memory
  
• Added better error handler for parameter declaration
  
• Added UCase function
  
• Added Lcase and fixed ucase to be case insensitive
  
• Added MultiKeys Function (similar to FREEBASIC) so multiple keys can be checked at once
  
• Added GetKeyScanCodes
  
• Added HEX and HEX16 functions to return HEXadecimal string representation of numbers
  
• Fixed a bug when a$(n TO) was specified
  
• Optimization: Remove unnecessary jumps at function returns.
  
• STORE16 now generates a more efficient backend (Z80 ASM) code.
  
• Added alias for arrays. Now you can declare:
  
  Code:

  DIM a(10)
  DIM c at @a
• Better code generation for STORE32 and STOREF backend
  
• Optimized constant array assignation as a direct store.
  
• Added constant array Read access optimizations
  

This release includes new fixes and improvements:

• Fixed a bug in the PRINT command when printing floating point numbers due to a Spectrum ROM Bug. The temporary attributer were not being used.
  
• Fixed a bug in the STR$ function that made the program to freeze and also clears the temporary attributes (related to the ROM bug as above).
  
• Fixed a bug in the assembler which, sometimes won't assemble if #init directives were used. Also some code clean up.
  
• Now you can enter ink/paper codes and graphic characters ala BASIN (read below)
  
• Added shift bit instructions (SHL, SHR)
  

This release includes new fixes and improvements:

• Fixes a bug in the DRAW command. The draw implementation was buggy. The new DRAW implementation works ok, and is faster and smaller.
  
• Support for DRAW with 3 arguments. Now you can DRAW arcs as in Sinclair BASIC. The routine works exactly as the ZX Spectrum ROM, but also allows drawing in the lower 16 scan lines
  
• Optimization: If a function is not called, and you use -O1 (or a higher number), it won't be compiled, allowing saving space
  
• Optimization: If a global variable is not used, and you use -O2 (or a higher number), it won't be compiled, allowing saving space
  
• Optimization: The code generated for temporary atributes is now smaller and faster
  

• Use --array-base=1 to start arrays at subscript 1 (default is 0)
  
• Use -On to enable optimizations. Higher n number leads to higher optimizations
  

Defining constants with CONST Constant as UByte=0 is not the ideal way, because they can be easy mistaken for variables, and can interfer with variables. So my proposal is to change it to similar way, PureBasic does:
#Constant as UByte=0, and all constants are marked with "#" sign in the front. In PureBasic it is a little different, as the constant and variable types are defined this way: #constant.ub=0 (ub=short for unsigned byte, has to be defined only one time).
apart from this I propose to include build-in constants with ROM Labels and System variables:
#KSTATE=23552, #LASTK=23560, #REPDEL=23561, #REPPER=23562, ...
#start=$0, #error_1=$8, #print_a_1=$10, #get_char=$18
#test_char=$1C, #next_char=$20, #fp_calc=$28, #bc_spaces=$30, ...
Also constants for some other usages:
#False=0, #True=1, #Zero=0
#Black=0, #Blue=1, #Red=2, #Magenta=3,...

Is this a good idea???

There is (already) a Wiki with the online documentation.

Some users are unaware of this documentation. Please, read there before asking for help since the documentation includes a programming reference and lot of examples, code snippets, etc.

Note: To contribute to the wiki you must register (create an account) in this forum, and later login into the wiki with the same user/pass.

New version 1.0.5 released.

This version fixes very important bugs. Please download this version before reporting problems with the compiler.