'(float)"-nan"' doesn't return Indet in LSL:
llOwnerSay(llList2CSV([(float)"-nan"])); // outputs "nan", not "-nan"
Therefore, for the output to yield the correct result we have to use a different strategy to generate an indeterminate. We choose '(1e40*0)' which is shorter than the rest.
Also, we don't output infinites as '(float)"[-]inf"' but alwas as '1e40' or '(float)-1e40' (or just '-1e40' if we're in globals).
Commit 5804a9a introduced a bug where having the foldtabs option disabled (normal) prevented optimizations of functions. Fix it for good (hopefully). While on it, rename the nofoldtabs option to warntabs, making it default, and use it to disable a warning when there are tabs in a string.
Adds a new tree node type, SUBIDX, which hopefully should never appear in actual output. If it does, it's prefixed with the string (MISSING TYPE) as a cue to the programmer.
The output module adds parentheses where necessary, depending on the evaluation order in the tree. Or that's the idea. Prone to bugs, let's see how it bodes.
There's also a new hidden option, shrinknames, which automatically enables duplabels due to its nature. The idea is that once general renaming is implemented, in order for label names to not cause collision trouble, they are renamed out of the way with unique names.
Not entirely sure this is really necessary.
Fixed by backtracking in the parser, and keeping a copy of the original expression if it's a simple_expr, which is used for output in place of the folded one.
There's still the potential issue that if a global is optimized away, then it will "come back" during output and cause an error because the definition is missing.
Bugs fixed:
- %= and the new assignment operators were not emitting error on invalid types.
- List globals referenced in another global were duplicated entirely.
- Properly recognize -option in the command line.
Rest:
- Complete overhaul of the internal data structure.
- Got rid of the symbol table plus mini-trees, and made everything one big tree plus an auxiliary symbol table.
- No more special case hacks like using tuples instead of lists...
- Got rid of the EXPR hack.
- Dict-based, rather than list-based. Allows adding arbitrary data to any node or symbol entry.
- Added a few coverage tests for the new code.
- Return values can now be chained; the functions parameter requirement is gone. Still not fully convinced, though. My guess is that a parser object should be passed between functions instead. Will do for now.
- Get rid of Fold().
- Handle globalmode properly. It was sometimes active during function calls.
- Change all warning() calls to not use Unicode, just in case the output is redirected to file.
- Cosmetic fixes and TODO items.
- Parser and output modules are thoroughly tested and working.
- Most LSL immutable functions are working; some not tested; llJsonSetValue not implemented.
- Parser recognizes the following flags that alter syntax:
extendedglobalexpr: Allow full expression syntax in globals.
extendedtypecast: Allow full unary expressions in typecasts e.g. (float)~i.
extendedassignment: Enable the C assignment operators &=, ^=, |=, <<=, >>=.
explicitcast: Add explicit casts wherever they are done implicitly, e.g. float f=3; -> float f=(float)3;.
Of them, only extendedglobalexpr is useless so far, as it requires the optimizer to be working.
