Since MatlabParser has no reason to have an instance of it managed by a
user, just make it use a single static method to access its functionality.
Overall it makes code simpler (avoiding the nasty getRootExpression call).

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10847 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Instead of having new instances of VariableParser created, and thus wrapping
a Map<String, Variable> in a very inelegant way, simply have a single method
that actually does the VariableParser work, and directly returns the Map.
Also ensure all tests + Matlab continue to work, and tighten up some extra
imports and Matlab output in affected areas.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10846 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The package name matlab2smt is really ugly, and not really representative of
what it really does anymore.  Thus create an expression package for the pieces
representing an expression, and a parsers package for all parsers, including
both the variable and the Matlab parsers.  Also make
MatlabParserOperatorParseTest not rely on (now) private members.  Instead
use the toString methods to get a generator independent representation, and
test that it matches appropriately.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10845 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Since the java code is turning into entire toolbox for dealing with tabular
expressions, move the code to an appropriate package.  Even if it doesn't
become a full toolbox (for java), it is deeply used by the Matlab toolbox.
Thus pull the name from there.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10844 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Add some toString methods to aid debugging the operations setup.  It isn't
committed centrally as I'm not sure it is a good idea.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10843 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Handle the negative and fractional outputs from SMTLIB.  This required
extending the SMTLIB output grammar to accept negative and division operations.
Currently the support is cheated some, so that only the necessary pieces are
actually parsed.  Further support for more operations will require extending
the support further.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10724 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Since periods are usually a wildcard, make sure to quote it so that it is
properly handled as a period, and not any random value.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10723 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

To handle the case when more complicated output is received from SMTLIB,
properly parse SMTLIB output.

Due to ANTLR trying to consume the full stream, the SMTLIB response is first
put in a string.  A simple algorithm, using the brackets, is used to find the
beginning and end.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10722 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Update antlr to the latest release.  Also adjust the Matlab Grammar to continue
working.  No functional change necessary.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10721 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Update the assembly plugin, used to generate the jar with dependencies, to
the latest released (non-beta) version.  No functional change.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10720 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

CVC3 doesn't allow decimal points to represent rational numbers.  Thus simplify
things by putting the literal into fractional form.  Currently reduction is
not attempted, and the simplest form is used.  Ex. 4.56 becomes (456/100) (as
tested in the accompanying unit test).

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10714 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Since queries is now a Java List, don't use matlab to combine them (it makes
an array of Lists).  Instead use the Java method addAll to do it.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10713 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The CVC3 generator originally insured that the TCC numbering was always unqie
for a table.  Re-add this functionality.  Specifically, when building the walker,
add a parameter to its constructor to set its starting TCC number.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10712 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Fixed point operations are now all tested.  Due to literals automatically
casting to reals, it is necessary to push the "x" variable further into the
expression.  To ensure everything works, add one more condition without x to
make sure everything stays stable.  Everything works, after adjusting search
strings.

Also, deal with both signed/unsigned fixed point types.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10711 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

When generating binary operations, the requested type sent down was the output
type, not the input type.  However, the generator expects the input type to
know how to output the operation.  Fix that.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10710 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Add a test for CVC3 fixed point support, and get fixed point working again.
This implements support for the fixed point marker, and make use of it.

The only problem is that literals by default fall down to Reals, so not all
operations are tested.  This needs to be fixed.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10709 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Now that Matlab parsing is pretty much all done, remove extra output.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10708 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Real operators are now added for CVC3 generation.  Fixed point is still left
out.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10707 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Implement other operators for SMTLIB.  Also implement a test for it.  The test
is parameterized, so CVC3 support is easy to add.  Due to the parameterization,
a newer version of JUnit is required, so bump the version in the pom.xml file.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10706 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

When dealing with expressions, and generating output for them, types have
to be exactly known.  However, when two literals are added together (like:
5+5), the type to use for that is unknown.  To fix this issue, type information
has to be forced down the pipe.  So know the addition operation now knows it
needs to produce a particular type (ex Real or Fixed Point), and then tests
what the subexpressions can produce (for literals, they can produce either,
but a variable can only be what it is).

For Real, that is fine.  However a fixed point can take an infinite number of
different possibilities, and one cannot test for every combination.  So instead
a type "marker" is introduced.  This marker instead identifies that a fixed
point type is generally wanted, but specifics are unknown.  The system will
query for an appropriate type, and requires the specific type is returned, and
not a marker.  All types are also markers for themselves, so Reals, for instance,
don't need anything special.

While the marker system sounds good in theory, it currently is fully implemented,
so fixed point is currently broken.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10705 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Binary ops now includes:
-
/
>=
<
<=
Unary:
-
YAY!

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10704 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Test that each operator is parsed in properly.  Currently only tests implemented
operators, with more to follow.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10703 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The java code now loads into Matlab automatically, but it uses a differnt
jar file location now.  Things work though.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10668 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Show an option for checking against Z3, and also hook it up to the matlab
code for use.  Everything should be working now.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10667 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Do to the way the matlab code works, the GenerateMatlabVariableDeclarations
is actually useless.  Thus, make SampleValues public so that the matlab code
can still do what it needs to do.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10666 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

When walking an empty grid, no queries for either complete or disjoint are
generated.  Disjoint handled this fine, but complete did not.  Fix this.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10665 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Due to some copy+paste errors, cvc3 was mentioned in smtlib test code.  Rename
the references as appropriate.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10664 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Add a CheckerRunner class with a single static method to run z3.  It uses
the smtlib generators, and the logic can be extended to other smtlib2 compliant
smt solvers.  It uses the copy of z3 in your path for now.

Note that it is very brittle, and if z3 has issue can potentially lock the
thread it is running on, as Java has no easy way to add timeouts to reads
on the stdout of z3.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10663 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Add a simple class to hold the results of a failed query to a SMT checker.
Also, it has a utility function to easily generate a matlab expression for
the failed test's counter-model.  Also test.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10662 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The grammar file requires Matlab expressions, it expects the expression to end
with a newline.  However, lots of generated code doesn't.  Thus force this to be
always correct.  Note that any extra newlines shouldn't cause issues.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10661 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

To stay at least a little consistent, use smtlibchecker for the smtlib
package name.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10660 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The cvc3generator package was moved out of tablularexpression.  Matlab code
wasn't updated, and thus broke.  Fix this mistake.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10659 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Implement a first run (with simple non-exhaustive test) of the SMT-LIB generator
for hierarchical tables.  This generates the same information as the cvc3
generator, but using the appropriate syntax.  Note that the generated script
can't be directly run, due to how SMT-LIB verifiers work regarding reading of
values, so the queries need to be piped directly across one by one for
verification.

Also, since SMT-LIB doesn't have an equivalent to CVC3's QUERY command, all
queries have a not placed in front.  This lead to a few changes to the queries:
 - Disjoint queries now had a format of: (not (and (not (and e1 e2)))).  This
can be simplified with De Morgan's law to (or (and e1 e2)) (by removing the
(not (and (not to the more simple (or).
 - Complete queries now have an extra (not in front.  No further optimization
is possible.
 - For the parent part of any query, instead of using (not (=> (p e)), which
removes the potential disjoint optimization, the not is pushed down so the
format is now (and p (not e)).  Also, for simplification p is always defined,
and is true when no parent cells exist.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10658 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

In SMT-LIB, any real literal *must* have a decimal point, otherwise various
parsers may explode.  Since I know if a literal is supposed to be real, fix
this issue by appending a harmless ".0" to the string of the number.  Fix
tests to verify this.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10657 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The system can now generate SMT-LIB code from the parsed Matlab code.  It doesn't
deal with tables (yet), but that is next.  Also use the MatlabParser tests to
verify the generated SMT-LIB code.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10656 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Start with the SMTLIBGenerator class.  To start with, deal with variable
definitions!

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10655 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

GenerateVariableType was an unused function outside of some internal pieces.
Thus remove it from the CheckerGenerator interface, as it doesn't need to be
exposed.  Update a few internal uses of it to use an explicit instance of a
CheckerGenerator for when it was used (inside of tests).

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10654 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Some debug output was left in and committed.  Remove the extra junk.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10653 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

The cvc3generator package was originally put in tablularexpression while
I was fighting with maven.  Since it doesn't have to be in the package, remove
it to clean up the layout of packages.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10652 57e6efec-57d4-0310-aeb1-a6c144bb1a8b

Now the full contents of the CVC3 file are generated inside of Java, but the
actual saving and running of CVC3 still happen inside of matlab.  For now.

git-svn-id: https://groke.mcmaster.ca/svn/grad/colin/branches/TableTool_javization@10651 57e6efec-57d4-0310-aeb1-a6c144bb1a8b