Commit graph

26 commits

Author SHA1 Message Date
941fe3a9b9 factor out util function to "stringify" arbitrary arguments
this function is of use also for creating a vector of strings
from a bunch of C-Strings, but it could also be used to
construct other stuff initialised by strings (e.g. RegExps)
2015-12-12 23:18:25 +01:00
33f7fe116a implement refinement filter on log entry's arguments.
Whew! functional programming is such a powerful concept.
You get additional refinement and lazy backtracking
basically for free....
2015-12-12 03:13:32 +01:00
5bc6919bdb implement the specialised event or call-matching logic 2015-12-12 02:23:12 +01:00
80ffdc1f05 provide the remaining, specialised logging calls 2015-12-12 01:35:38 +01:00
261bcc9fdb implement logging of "events" with additional ID / classifier 2015-12-12 01:28:42 +01:00
4522e209fb implement generic log entry with arbitrary fields 2015-12-12 01:24:52 +01:00
894ef68a8f EventLog: implement logging of function invocations 2015-12-12 01:01:46 +01:00
761154ae63 stub the code into submission.... 2015-12-11 20:02:30 +01:00
bf92333339 implement log joining in shared heap storage 2015-12-09 01:18:15 +01:00
cd8e6d874c EventLog: switch to shared PImpl
this is prerequisite for joining and sharing logs
2015-12-08 22:14:29 +01:00
09afbb0e12 change implementation technique: use flags instead of exceptions
abandon the use of an assertion exception to signal match failure,
rather use a final bool conversion to retrieve the results.

Error messages are now delivered by side effect into STDERR


The reason is we're unable to deliver the desisred behaviour
with the chosen DSL syntax in C++ ; on a second thought the
new approach is even better aligned with the overall way
we're writing tests in Lumiera. And we produce match-trace
messages to indicate the complete matching path now
2015-12-08 03:20:52 +01:00
b3881696e5 this solution for negative matches is not usable
implemented a solution to determine negative matches.
But because this solution relies on throwing from a destructor,
it is not possible to catch the resulting assertion failure.
Not sure why (AFAIK there is no second exception thrown
while unwinding the stack), but throwing from dtors is
considered "undefined behaviour" anyway.

So this solution is of limited use


beyond that solution, I'm not sure if the desired syntax
can be implemented at all in C++. Seems that we need to build
a bracketing construct, first to initiate a negated match
and finally, after all queries, to detect if there happened
any failure or not
2015-12-08 01:10:02 +01:00
272d62d5a3 log handling const correctness
adding log entries requires full access,
whereas matching log entries is const,
since it doesn't alter the log
2015-12-06 04:37:41 +01:00
a95f9a6cac draft how "log joining" shall work 2015-12-06 04:21:35 +01:00
c9d113be87 EventLog: implement match backwards
..based on the new IterCursor, which gives us the ability
to switch the iteration direction
2015-12-06 03:35:07 +01:00
00abf9f1f9 err: got the naming and the junctor condition backwards
the junctor is called "before" but searches ahead.
And in this case we do not need to *extend* the filter condition,
just replace it with a new one...
2015-12-05 03:37:25 +01:00
09a560dfac add explicit diagnostics to assertion failure
...and fix an error (header include order of diagnostics facility)
which prevented the first matcher implementation to work

the after()-match now works as expected
2015-12-05 03:13:01 +01:00
9300a0baaf code up the first event log matching predicate
this establishes the basic pattern used for implementation
Note: test still fails
2015-12-05 02:40:03 +01:00
38fd7d6d69 use the new extensible filter iterator to implement event log matching 2015-12-05 02:07:25 +01:00
68dd8a9e03 refine implementation draft: make FilterIter extensible on-the-fly
after looking into our various iterator tools,
it seems obvious that our filtering iterator implementation
has almost all of the required behaviour; we only need to
add a hook to rewrite and extend the filtering functor,
which can now nicely done with a lambda closure.

This means all memory management, if necessary, is
pushed into std::function and the automated memory
management for closures provided by the runtime.
2015-12-05 00:28:07 +01:00
41626c363e WIP: draft how to implement those chained EventLog matchers
the solver is a Monad; succesively flatMap more conditions
and try to get some element from the solution
2015-12-02 23:37:54 +01:00
d38b28da5b implement initial EventLog entry 2015-12-02 01:31:37 +01:00
b2542b86f7 stub and rectify interfaces defined thus far.
...compilation PASS again
2015-11-28 23:50:56 +01:00
809ed36b56 WIP: draft initial test for event logging helper 2015-11-28 19:20:10 +01:00
1eda2a070b harmonise the form of the header include guards
no trailing underscore
2015-11-28 18:36:35 +01:00
d04e6d74d8 WIP: arrange some elements needed for MockElm implementation 2015-11-27 19:24:00 +01:00