Re: Ann: SWI-Prolog 6.5.1

Hi Jan,
Thanks for the new version - I have a few versions to catch up now,
but if you're wondering what to do next, might I suggest …
 [whispering] tabling?
I'd love to be able to implement something like this
	http://sato-www.cs.titech.ac.jp/prism/
on SWI. BProlog seems like a good Prolog but in terms of general
usability, not to mention my investment in libraries, SWI is a lot better
for me. I realise adding any sort of efficient tabling probably involves a major 
reworking of the base code - is it at all doable?

Samer.

On 12 Aug 2013, at 21:43, Jan Wielemaker <[email protected]>
 wrote:

> Hi,
> 
> I've uploaded SWI-Prolog 6.5.1.  There are a lot of changes.  Highlights:
> 
>  - Detection of `semantic singletons' by the compiler.  This proofs
>    to result in a quite high percentage of real bugs.  My rough
>    estimate is about 1/3th!  And this is on established code!
>  - Singleton detection with quasi quotations (do not claim a
>    singleton of a QQ introduces a reference to the singleton
>    in its expansion).
>  - prolog_load_context/2 provides access to variable names.
>  - term_expansion/4 and goal_expansion/4 to allow converting
>    source layout information.  This goes through most of the
>    compiler toplevel, but isn't actually used yet.
>  - halt/0 can now be called from threads.  Various fixes to
>    program termination, notably for xpce on Windows.
>  - JPL dll hassle is hopefully resolved.
>  - Cygwin port is updated.
>  - Lots of small stuff and cleanup.
> 
> Not sure where to go with the stable version.  I first thought just
> cherry-picking the clear fixes, but there are so many that I consider
> updating to the devel version after the devel version proofs to be
> stable.  Please report issues with the above changes.
> 
> 	Enjoy --- Jan
> 
> 
> ========================================
> SWI-Prolog Changelog since version 6.5.0
> ========================================
> 
> * DOC: keysort/2
> 
> * ENHANCED: CLP(FD): isolate global_cardinality/2 from user attributes.
>   Before:
> 
>      %?- put_attr(A, edges, [a]), global_cardinality([A,B,C], [0-_,1-_]).
>      %@ false.
> 
>   Now:
> 
>      %?- put_attr(A, edges, [a]), global_cardinality([A,B,C], [0-_,1-_]).
>      %@ put_attr(A, edges, [a]),
>      %@ A in 0..1,
>      %@ global_cardinality([A, B, C], [0-_G4383, 1-_G4389]),
>      %@ B in 0..1,
>      %@ _G4383 in 0..3,
>      %@ C in 0..1,
>      %@ _G4389 in 0..3.
> 
> * FIXED: Dummy empty warning if no compiler warning refers to a variable.
> 
> * ENHANCED: CLP(FD): Better isolation from user-defined variable
>   attributes.  Example:
> 
>      :- use_module(library(clpfd)).
> 
>      distinct(Vars) :- Vars = [A,B,C,D], all_distinct(Vars), [A,B,C]
>              ins 0..2.
> 
>   Before:
> 
>      %?- put_attr(X, edges, [x]), Vars = [X|_], distinct(Vars).
>      %@ false.
> 
>   Now:
> 
>      %?- put_attr(X, edges, [x]), Vars = [X|_], distinct(Vars).
>      %@ Vars = [X, _G1856, _G1859, _G1862],
>      %@ put_attr(X, edges, [x]),
>      %@ X in 0..2,
>      %@ all_distinct([X, _G1856, _G1859, _G1862]),
>      %@ _G1856 in 0..2,
>      %@ _G1859 in 0..2.
> 
>   That is, users can freely use the "edges" attribute themselves.
> 
> * ENHANCED: CLP(FD): Remove branch singletons.
> 
> * FIXED: Inlined unification must revert to normal unification if occurs
>              check is active.
> 
> * TEST: compensate for optimization
> 
> * FIXED: Mixup in truth of == and \== predictable warnings.  Keri Harris.
> 
> * ADDED: Move multiton detection into the compiler, so this compiles
>     clean: == t :- (v(_A);v(_A)).  ==
> 
> * FIXED: Detection of `multitons' (broken after singleton detection
>     extension for dealing with quasi quotations).
> 
> * FIXED: myplus/4 in bounds.pl.  Note that bounds is deprecated in
>     favour or library(clpfd).  Markus Triska.
> 
> * ENHANCED: :- at_halt/1 is now term-expanded, such that reloading the
>     file does not re-register the hook.
> 
> * FIXED: Apply changes that where missed due to a git conflict
> 
> * UPDATED: Handle new PL_on_halt() interface
> 
> * MODIFIED: Program termination: - at_halt/1 handlers can veto
>       termination of halt/0 by calling
>         cancel_halt/1.  - PL_at_halt() handlers now have a declared
>       return type.  In
>         the future, they will be able to cancel termination by
>         returning a non-zero status.  Now a warning is printed.  -
>       halt/1 and PL_halt() can be called from any thread.
> 
> * FIXED: set_stream/2 issues with recent patch
> 
> * ENHANCED: safe_goal/1, various enhancements: - Distinguish undefined
>     predicates - Deal with library meta-predicates (security fix) -
>     Module aware call/N support - Print context of encountered problem
> 
> * CLEANUP: Use DCG // notation for declarations
> 
> * FIXED: license/1: list known licenses (found by singleton analysis)
> 
> * ENHANCED: read_term/3: singleton detection with quasi
>     quotations. Variables appearing only ones are searched in the quasi
>     quotation results.  If found there, they are no real singletons.
> 
> * ADDED: library(sandbox): added a lot of safe primitives.
> 
> * FIXED: safe_goal/1: Catch instantiation_error when trying to proof
>     safety of a general goal.
> 
> * MODIFIED: Make set_stream/2 behave sensibly on stream pairs. Daniël
>     de Kok.
> 
> * CLEANUP: Implementation of set_stream/2.
> 
> * ADDED: Singleton checking inside \+
> 
> * DOC: var_branches style check
> 
> * CLEANUP: Removed two useless assignments
> 
> * FIXED: nb rbtrees (singleton)
> 
> * MODIFIED: Setting style_check/1 and expects_dialect/1 in ~/.plrc
>     changes the default for the session.  The settings used to be
>     scoped to the file.  This scoping is now controlled by a new option
>     =scope_settings= to load_files/2.
> 
> * ADDED: style_check/1: ?- style_check(?(Style)). to return active
>     style checks on backtracking.
> 
> * DOC: term_expansion/4, etc.
> 
> * DOC: Document extended singleton warnings
> 
> * DOC: style_check/1 new =no_effect= option
> 
> * DOC: user term_expansion is followed by DCG!
> 
> * CLEANUP: Avoid new warnings in test cases.  Test cases sometimes use
>     ugly code by design ...
> 
> * ADDED: compiler warnings for BIPS that have no effect and singleton
>     variables in branches (semantic singletons).
> 
> * ADDED: Support for warning messages from the compiler
> 
> * ADDED: Provide prolog_load_context/2 =variable_names= and implemented
>     goal/term expansion with 4 arguments to process layout information.
> 
> * FIXED: Default handling in setting (spotted by variable balance
>     checking)
> 
> * CLEANUP: some documentation glitches
> 
> * ADDED: New style-check options.  We will include these into the stable
>     branch to avoid warnings on code calling style_check/1.
> 
> * TEST: Fixed test code for timeout handling.
> 
> * CLEANUP: Removed bogus test
> 
> * FIXED: Remove meta_predicate/1 declaration for '|'/2.  This is since
>     long not a predicate anymore.
> 
> * CLEANUP: Use '$type_error'/2.
> 
> * FIXED: Colourise [:|+] meta-declaration.
> 
> * CLEANUP: Rename $-predicates in DCG expander. This is not needed since
>     this code is in a module. Also removed a lot of outdated comments
>     and added some new ones.
> 
> * FIXED: base32//1: Singleton I3
> 
> * CLEANUP: library(clpp/bounds): singletons.  Not sure whether this
>     was intended.
> 
> * CLEANUP: Branch singletons
> 
> * CLEANUP: Unification against unused variable
> 
> * CLEANUP: In DCG, \+ \+ {...} leads to useless unifications.  We now
>     use { \+ \+ G }.  That remains silent.  Eventually, this can (and
>     should) be improved in the DCG compiler.
> 
> * FIXED: Emulated shell commands for non-atom arguments.
> 
> * FIXED: library(prolog_pack): parsing of versions with * in them.
> 
> * FIXED: Pass thread-reference in all branches
> 
> * FIXED: library(prolog_colour) to colour "..." terminals properly
> 
> * PORT: Cygwin: install package shared objects in the right
>     location. Corinna Vinschen.
> 
> * PORT: Updated Cygwin port by Corinna Vinschen
> 
> * FIXED: make/0: avoid loading files twice.  Paulo Moura.
> 
> * FIXED: PL_thread_engine_attach() did not initialise the thread
>     locale. Carlo Capelli.
> 
> * FIXED: Make sure the config var SO_PATH is set to PATH in Windows,
>     and not WINEPATH.
> 
> * CLEANUP: Avoid using SO_PATH for two different purposes
> 
> * FIXED: DCG argument of phrase_from_stream/2 is not identified as such
>     (meta_predicate/1).
> 
> 
> 
> =========
> Package R
> =========
> 
> * FIXED: Mixup of Ruo/Ro variables in r_lines/6
> 
> ===============
> Package archive
> ===============
> 
> * PORT: Deal Win64 off_t handling
> 
> ===========
> Package chr
> ===========
> 
> * ADDED: style_check(-no_effect) to compensate for some unifications
>   that have no effect.  Suggests that is some room for optimization.
> 
> * CLEANUP: Avoid (extended) singleton warnings.
> 
> * FIXED: delete_first_ht/3: Not only compute load, but also update it.
> 
> ============
> Package clib
> ============
> 
> * UPDATED: Handle new PL_on_halt() interface
> 
> * PORT: Cygwin updates by Corinna Vinschen
> 
> ===========
> Package cpp
> ===========
> 
> * DOC: Document PREDICATE0() and NAMED_PREDICATE()
> 
> * ADDED: PREDICATE0(name) as alternative to PREDICATE(name, 0) that
>   avoids unused parameter warnings.
> 
> * MODIFIED: NAMED_PREDICATE() takes a string as name.
> 
> * FIXED: PREDICATE() macro issues
> 
> * ADDED: Macros NAMED_PREDICATE() and NAMED_PREDICATE_NONDET() to
>   allow defining predicates that have a name that cannot be expressed
>   as a C identifier.  This patch also casts unused variables to (void)
>   to avoid warnings.  With help from Carlo Capelli.
> 
> ============
> Package http
> ============
> 
> * CLEANUP: More elegant implementation for computing byte-length
>   in encoding
> 
> * CLEANUP: Branch singletons
> 
> * CLEANUP: Another branch singleton.
> 
> * FIXED: Singleton in interactive_httpd->authorise causes poor error
>   message.
> 
> * CLEANUP: Removed detection of unused OpenID NS
> 
> * INSTALL: Avoid double installation of shared objects. Corinna Vinschen.
> 
> * DOC: Improved documentation of cookie handling
> 
> 
> ===========
> Package jpl
> ===========
> 
> * PORT: Use win_add_dll_directory/1 and generalise the search a bit
> 
> =============
> Package pldoc
> =============
> 
> * CLEANUP: Avoid semantic singleton message
> 
> ==============
> Package plunit
> ==============
> 
> * FIXED: Only share variables between options and body to maintain
>   first-var and singleton properties.
> 
> * CLEANUP: Singletons
> 
> ==============
> Package semweb
> ==============
> 
> * FIXED: Avoid unlocking the RDF store directory twice.
> 
> * FIXED: rdfe_undo (bug found by balanced var check)
> 
> * FIXED: Loading Turtle from a stream did not reset the stream encoding.
> 
> * FIXED: Detecting subjects for turtle writer.
> 
> =============
> Package space
> =============
> 
> * FIXED: sem_time_candidate
> 
> * FIXED: owl_time_xsd_candidate/3 error spotted by singleton detection.
> 
> ===============
> Package windows
> ===============
> 
> * CLEANUP: Do not ignore return values and delete dead code.
> 
> ============
> Package xpce
> ============
> 
> * FIXED: Locking issue that causes long delays on halt if xpce is
>   running in separate thread.
> 
> * FIXED: Two new singleton warnings (one was a bug).
> 
> * FIXED: Priority issue in GUI term pretty printer
> 
> * UPDATED: Handle new PL_on_halt() interface
> 
> * ENHANCED: Allow cancelling halt
> 
> * CLEANUP: Variable style issues.
> 
> * PORT: Rename installation helper xpce-install into xpce-copy to prevent
>   Windows from trying to run this program as admin.  Corinna Vinshen.
> 
> * FIXED: Start xpce from a background thread.
> 
> * PORT: update of the Cygwin port by Corinna Vinschen
> 
> * CLEANUP: Avoid more semantic singletons
> 
> * FIXED: Semantic singletons.  One fixes a bug in the class
>   prolog_predicate.
> _______________________________________________
> SWI-Prolog mailing list
> [email protected]
> https://lists.iai.uni-bonn.de/mailman/listinfo.cgi/swi-prolog