Remove the multiple caching tables we were using
and use the grammar table for caching. This works
well because we can use raw_get for checking the local cache, and normal
get fro checking the global cache.
A keyword reference only counts as visited if we have
it as cached in the memoized->table, and we know it was
originally referenced from the same grammar table. If these
two conditions are true, then compilation must work correctly.
Also add janet_table_get_ex.
(backmatch [tag?]) is similar to a back reference in regular expressions
(NOT to backwards capture in a peg). It only matches a pattern if
it exactly matches the text of the last capture. It does not consume
or push any captures to the capture stack.
This will allow some one constructing an abstract to
only make it visible to the garbage collector after it
is in a valid state. If code in the constructing cfunction
panics before janet_abstract_end is called, the GC will not try
to mark the incomplete abstract type. This is often not needed through
careful programming, but should work well.
By holding on a reference to argv for a long time, we
may trigger a use after free bug if the stack is resized. In
janet c function, argv is only vvalid up until the next stack operation
on the fiber. We could say that this is the dynamic lifetime of
argv.
To fix this, we copy extra arguments into a tuple, which is properly
garbage collected.
Some peg grammars could not capture values based on their position in a
larger grammar. This is a design limitation inheritted from LPeg, but no
longer needed as the replace mode is superseded by the accumulator mode,
which is more general if slightly harder to use.
A consistent style should help with contributors and
readability. We use astyle as the formatter as can make a pretty
good approximation of the current style and my preferred style.
Astyle can be found at http://astyle.sourceforge.net/astyle.html
This should speed up start time and reduce malloc/free
usage to about 15% of what is what previously for startup.
The current cost is slightly larger binary as the representaion
of the image is currently less compact than source code.
After we changed peg bytecode emission to
preallocate space for an instruction before
emitting sub rules, the rules are numbered
in the order that they are compiled. This means
that the main rule is always 0.
We can remove the explicitly stored main rule in
the peg structure.
Disable tail calls in the root scope for better
stacktraces, as the root scope may contain a single call
to a failing function, as in the case of the test suite.
This allows that pattern to call an external function to
check if some text should match or not. This allows for
matching any possible language a computer can recognize.
to make janet function calls easier and faster from C (still
needs an object pool for fibers, though). Fix bug in scan-number
and add many more peg tests.
as back references for PEGs. More documentation is needed for PEG
syntax, but the amount required will need an external document, not
just a docstring.
