February 2002 - Cvs-nhc98

Updated Hat user manual.
by olaf＠sparud.net 07 Feb '02

07 Feb '02

olaf: Thu Feb 7 20:34:49 CET 2002 Update of /usr/src/master/nhc/docs/hat In directory hinken:/tmp/cvs-serv10954 Modified Files: hatuser.tex Log Message: Updated Hat user manual.

1 0

Improvements to hat-explore.
by malcolm＠sparud.net 07 Feb '02

07 Feb '02

malcolm: Thu Feb 7 18:27:09 CET 2002 Update of /usr/src/master/nhc/src/hat/tools In directory hinken:/tmp/cvs-serv10105 Modified Files: HatExplore.hs Makefile SExp.hs utils.c utils.h Added Files: LowLevel.hs Log Message: Improvements to hat-explore. * Use a different low-level interface which gives correct parent information for identifiers, constructors, and literals. * In the state, store a stack of sub-expression contexts, not just expressions. Thus, when you go backwards up the display, the cursor returns to the previously selected subexpression, not the whole line.

1 0

Adjust top-level configure and Makefile for new tool `hat-
by malcolm＠sparud.net 06 Feb '02

06 Feb '02

malcolm: Wed Feb 6 18:16:55 CET 2002 Update of /usr/src/master/nhc In directory hinken:/tmp/cvs-serv27237 Modified Files: configure Makefile Log Message: Adjust top-level configure and Makefile for new tool `hat-explore', and bugfix version of `hmake' (3.01), and portable hatlib.

1 0

First version of `hat-explore', a textual version of hat-t
by malcolm＠sparud.net 06 Feb '02

06 Feb '02

malcolm: Wed Feb 6 18:15:19 CET 2002 Update of /usr/src/master/nhc/src/hat/tools In directory hinken:/tmp/cvs-serv27205 Modified Files: HatTrace.gc Makefile utils.c Added Files: SExp.hs HighlightStyle.hs PrettyLibHighlight.hs HatExplore.hs Log Message: First version of `hat-explore', a textual version of hat-trail.

1 0

Implemented labelled fields fully.
by olaf＠sparud.net 06 Feb '02

06 Feb '02

olaf: Wed Feb 6 13:28:55 CET 2002 Update of /usr/src/master/nhc/src/hat/lib In directory hinken:/tmp/cvs-serv25435/src/hat/lib Modified Files: Hat.hs Log Message: Implemented labelled fields fully. Data type definition, use of labels as selectors, construction and update with fields are now handled by hat-trans. Because there is no appropriate structure in the Hat trace file format, construction is encoded as normal construction of a data constructor and update is encoded as a "case". The latter can lead to weird expressions in hat-trail (e.g. "f main |> case C True undefined undefined"; "f (case C True undefined) undefined" would be clearer). Field labels only appear as selector functions in the trace. Selection and update can cause runtime errors if a field label is not present. To allow a simpler transformation, these runtime errors are not caught. A general solution to runtime errors is needed anyway.

1 0

Implemented labelled fields fully.
by olaf＠sparud.net 06 Feb '02

06 Feb '02

olaf: Wed Feb 6 13:28:54 CET 2002 Update of /usr/src/master/nhc/src/compiler98 In directory hinken:/tmp/cvs-serv25435/src/compiler98 Modified Files: Extra.hs Syntax.hs SyntaxPos.hs TraceTrans.hs Log Message: Implemented labelled fields fully. Data type definition, use of labels as selectors, construction and update with fields are now handled by hat-trans. Because there is no appropriate structure in the Hat trace file format, construction is encoded as normal construction of a data constructor and update is encoded as a "case". The latter can lead to weird expressions in hat-trail (e.g. "f main |> case C True undefined undefined"; "f (case C True undefined) undefined" would be clearer). Field labels only appear as selector functions in the trace. Selection and update can cause runtime errors if a field label is not present. To allow a simpler transformation, these runtime errors are not caught. A general solution to runtime errors is needed anyway.

1 0

Hack to avoid compiler error with -T.
by malcolm＠sparud.net 05 Feb '02

05 Feb '02

malcolm: Tue Feb 5 17:16:52 CET 2002 Update of /usr/src/master/nhc/src/prelude/IOExtras In directory hinken:/tmp/cvs-serv18401 Modified Files: DIORef.hs Log Message: Hack to avoid compiler error with -T.

1 0

* Re-implement IORef directly as a 1-item Vector, to elimi
by malcolm＠sparud.net 04 Feb '02

04 Feb '02

malcolm: Mon Feb 4 18:46:01 CET 2002 Update of /usr/src/master/nhc/src/prelude/IOExtras In directory hinken:/tmp/cvs-serv8590 Modified Files: DIORef.hs FreezeIOArray.hs IOExtras.hs NewIOArray.hs NewIORef.hs ReadIOArray.hs ReadIORef.hs WriteIOArray.hs WriteIORef.hs Log Message: * Re-implement IORef directly as a 1-item Vector, to eliminate the overhead of the Ix class in a 1-item IOArray. Should be somewhat faster now. * Remove '#ifdef 0' dead code from the IOArray operations.

1 0

Eliminate spurious warnings in hat-trans.
by malcolm＠sparud.net 04 Feb '02

04 Feb '02

malcolm: Mon Feb 4 16:58:54 CET 2002 Update of /usr/src/master/nhc/src/compiler98 In directory hinken:/tmp/cvs-serv7983 Modified Files: AuxFile.hs AuxFixity.hs AuxLabelAST.hs Log Message: Eliminate spurious warnings in hat-trans. * Previously, the IdentMap (that maps a constructor, field, or method to its type or class) was only used on entities defined in this module. This led to spurious warnings about some identifiers that were imported and re-exported. * Now, the IdentMap is always paired up with the AuxTree (mapping from Identifier to AuxInfo) to form an 'Environment', so the two are always carried together. Thus, getImports extends the IdentMap as well as the AuxTree. Also, when relabeling the abstract syntax tree, we now have correct information about the kind (Var, Con, Field, Method) of each entity, rather than a poor approximation.

1 0

Write traced reductions called from untraced code correctl
by olaf＠sparud.net 04 Feb '02

04 Feb '02

olaf: Mon Feb 4 15:29:28 CET 2002 Update of /usr/src/master/nhc/src/hat/lib In directory hinken:/tmp/cvs-serv7496/src/hat/lib Modified Files: Hat.hs Log Message: Write traced reductions called from untraced code correctly to trace. Example: filter myEven [1..10]; myEven x = x filter is untraced, but because myEven is traced the reductions myEven 1 = False, myEven 2 = True, ... should be in the trace. The "fun?" combinators responsible for this wrote the application but not the following Sat pointing to the result into the trace. After changing that it works. However, hat-detect cannot find such kind of children and hat-trail does not seem to have any trace pointing to them either; the trail is rather weird. Note that if myEven is defined as "myEven = even", then hat-observe myEven will only return myEven = even which is correct but not terribly useful.

1 0