Why does BridgePoint grammar allow ? as rval

[Bob Mulvey]

In discussion today, the BP oal grammar was discussed (QMARK in https://github.com/xtuml/bridgepoint/blob/master/src/org.xtuml.bp.als.oal/bnf/oal.bnf ). Specifically, there is support for an rval of ? (qmark in oal.bnf). Looking at the current implementation for this token I see no valid use case. Furthermore, there is inconsistency between the parser and verifier here (parser allows something that the verifier does not support).

Consider:

Example 1 (this IS currently valid syntax AND executes):
test=1;
test=?;
LOG::LogInteger(message:test);
// This displays the value 1

Example 2 (this is valid syntax, but execution fails on the first assignment because there is no type defined type in BPValue::getValueString):
test=?;
test=1;

I could see where example 2 could be useful to allow syntax where there was an unknown value or even unknown type that was initialized later, for example:

example 3 (not supported by the parser):
test=?
if (test==?)
test=1;
else
test=”string”; // parser error, type already defined
end if;

Does anyone understand the history of why this rval “?” is present?

[Lee Riemenschneider]

In this discussion, a workaround was shown for how to create an out-of-scope unassigned inst_ref variable. The QMARK seems to be what was put into OAL for this purpose with any type of data variable.

The problem with inst_ref is the QMARK variable can’t be used inside a select, so a workaround must still be used.
ref = ?;
if (true)
select object instance foo of BAR;
ref = foo;
end if;
if (empty ref)
end if;

Verifier should probably support the QMARK. MC-3020 seemed to handle it OK in an experiment I ran.

In a discussion yesterday, Levi and I surmised that the proper handling of no assignment should be a default value for the type.
e.g.,
ref = ?;
if (false)
select object instance foo of BAR;
ref = foo;
end if;
ref will be empty here due to no assignment.

• This reply was modified 6 years, 11 months ago by Lee Riemenschneider.

[Author]

Posts