OSDN Git Service

joypy/Thun.git
4 years agoMake sure joy.gui.main.__doc__ is at the top of the module.
Simon Forman [Fri, 24 Apr 2020 17:47:23 +0000 (10:47 -0700)]
Make sure joy.gui.main.__doc__ is at the top of the module.

4 years agoRemove `__next__`.
Simon Forman [Fri, 24 Apr 2020 17:32:48 +0000 (10:32 -0700)]
Remove `__next__`.

4 years agoAttributeError: 'itertools.count' object has no attribute '__next__'
Simon Forman [Fri, 24 Apr 2020 17:10:50 +0000 (10:10 -0700)]
AttributeError: 'itertools.count' object has no attribute '__next__'

4 years agoOpen stack pickle file in binary mode.
Simon Forman [Fri, 24 Apr 2020 17:10:30 +0000 (10:10 -0700)]
Open stack pickle file in binary mode.

4 years agoread text from REPL, don't eval
Simon Forman [Fri, 24 Apr 2020 07:03:32 +0000 (00:03 -0700)]
read text from REPL, don't eval

4 years agoPEP 518
Simon Forman [Fri, 24 Apr 2020 07:00:59 +0000 (00:00 -0700)]
PEP 518

4 years agoFake brutal_hackery on Py3.
Simon Forman [Fri, 24 Apr 2020 07:00:41 +0000 (00:00 -0700)]
Fake brutal_hackery on Py3.

Because I'm not about to try to figure it out tonight.

4 years agofuturize stage2 vui
Simon Forman [Fri, 24 Apr 2020 06:44:29 +0000 (23:44 -0700)]
futurize stage2 vui

4 years agofuturize stage2 gui
Simon Forman [Fri, 24 Apr 2020 06:38:10 +0000 (23:38 -0700)]
futurize stage2 gui

4 years agofuturize stage2 core
Simon Forman [Fri, 24 Apr 2020 06:33:23 +0000 (23:33 -0700)]
futurize stage2 core

4 years agofuturize stage2 core
Simon Forman [Fri, 24 Apr 2020 06:28:24 +0000 (23:28 -0700)]
futurize stage2 core

4 years agofuturize stage1 vui
Simon Forman [Fri, 24 Apr 2020 06:22:45 +0000 (23:22 -0700)]
futurize stage1 vui

4 years agofuturize stage1 gui
Simon Forman [Fri, 24 Apr 2020 06:19:29 +0000 (23:19 -0700)]
futurize stage1 gui

4 years agofuturize stage1 core
Simon Forman [Fri, 24 Apr 2020 06:16:45 +0000 (23:16 -0700)]
futurize stage1 core

4 years agoFixes #38239 rename parameter of stack.pick()
Simon Forman [Fri, 24 Apr 2020 04:39:29 +0000 (21:39 -0700)]
Fixes #38239  rename parameter of stack.pick()

4 years agoIgnore build and virtual env ("venv").
Simon Forman [Fri, 24 Apr 2020 01:59:12 +0000 (18:59 -0700)]
Ignore build and virtual env ("venv").

4 years agoBetter instuctions for running docs notebooks.
Simon Forman [Fri, 24 Apr 2020 01:53:32 +0000 (18:53 -0700)]
Better instuctions for running docs notebooks.

4 years agoHacked up the Prolog compiler code.
Simon Forman [Thu, 23 Apr 2020 22:58:38 +0000 (15:58 -0700)]
Hacked up the Prolog compiler code.

4 years agoDelete a pyc file.
Simon Forman [Thu, 23 Apr 2020 22:25:54 +0000 (15:25 -0700)]
Delete a pyc file.

4 years agoCrap, and stuff.
Simon Forman [Thu, 23 Apr 2020 22:18:11 +0000 (15:18 -0700)]
Crap, and stuff.

4 years agoShortcircuit Boolean combinators.
Simon Forman [Thu, 23 Apr 2020 22:15:36 +0000 (15:15 -0700)]
Shortcircuit Boolean combinators.

4 years agoSome comments.
Simon Forman [Thu, 23 Apr 2020 22:14:36 +0000 (15:14 -0700)]
Some comments.

4 years agoMinor edits.
Simon Forman [Fri, 31 Jan 2020 20:06:31 +0000 (12:06 -0800)]
Minor edits.

4 years agoMinor edits.
Simon Forman [Fri, 31 Jan 2020 20:01:13 +0000 (12:01 -0800)]
Minor edits.

4 years agoRemove unused predicates appears_only_once/2 and reg_used_once/2.
Simon Forman [Fri, 31 Jan 2020 16:30:10 +0000 (08:30 -0800)]
Remove unused predicates appears_only_once/2 and  reg_used_once/2.

4 years agoFreeing registers before using them is simpler.
Simon Forman [Fri, 31 Jan 2020 16:26:10 +0000 (08:26 -0800)]
Freeing registers before using them is simpler.

https://todo.sr.ht/~sforman/thun-der/3

4 years agoPartial reduction for combinator rule works after all.
Simon Forman [Thu, 30 Jan 2020 18:25:12 +0000 (10:25 -0800)]
Partial reduction for combinator rule works after all.

It just looked weird to me and I didn't think it through.
Once I checked it I realized it was okay.

4 years agoUncovered and fixed a subtle bug in free_reg//3.
Simon Forman [Wed, 29 Jan 2020 19:04:22 +0000 (11:04 -0800)]
Uncovered and fixed a subtle bug in free_reg//3.

non_alloc/1 for words that don't change the ref count of registers and can be delegated to their semantic relation.

4 years agoCons. I should research Lisp compilers...
Simon Forman [Wed, 29 Jan 2020 17:48:08 +0000 (09:48 -0800)]
Cons.  I should research Lisp compilers...

I think I should emit high-level code and reduce it to actual
machine code later under whatever model (cons cell heap, etc.)

4 years agoAdd assoc to kinda sorta track the values in the registers.
Simon Forman [Wed, 29 Jan 2020 02:04:51 +0000 (18:04 -0800)]
Add assoc to kinda sorta track the values in the registers.

But it doesn't update e.g. if you add two numbers, the value int(N) stays the same.

It could be modified to track the value as it computes?  But then why keep them in registers at all? Sometimes a value must arrive at runtime, eh?

4 years agoswap, pop, and +
Simon Forman [Wed, 29 Jan 2020 01:16:17 +0000 (17:16 -0800)]
swap, pop, and +

compiling is tricky

4 years agodup, add_ref/3.
Simon Forman [Tue, 28 Jan 2020 20:44:19 +0000 (12:44 -0800)]
dup, add_ref/3.

4 years agoTHread through a context to track registers.
Simon Forman [Tue, 28 Jan 2020 20:24:04 +0000 (12:24 -0800)]
THread through a context to track registers.

It seems to work to allocate and free registers with a kind of reference counting
by membership in an auxilliary list.

4 years agoSort of compile '+'.
Simon Forman [Tue, 28 Jan 2020 19:01:28 +0000 (11:01 -0800)]
Sort of compile '+'.

4 years agoMove immediate to register for int literal.
Simon Forman [Tue, 28 Jan 2020 18:46:38 +0000 (10:46 -0800)]
Move immediate to register for int literal.

4 years agoblep.
Simon Forman [Tue, 28 Jan 2020 18:37:13 +0000 (10:37 -0800)]
blep.

4 years agoA start on machine code generation.
Simon Forman [Tue, 28 Jan 2020 18:21:37 +0000 (10:21 -0800)]
A start on machine code generation.

Just the initial messing around...

4 years agoFix a bug in step.
Simon Forman [Mon, 27 Jan 2020 20:56:26 +0000 (12:56 -0800)]
Fix a bug in step.

You think it would be easy to find all the places where the type tags are needed.

4 years agodocs, formatter
Simon Forman [Mon, 27 Jan 2020 19:54:24 +0000 (11:54 -0800)]
docs, formatter

4 years agoFix a bug in rest.
Simon Forman [Mon, 27 Jan 2020 01:06:21 +0000 (17:06 -0800)]
Fix a bug in rest.

It didn't tag the tail list as a list.

Also, spell out bool for true cases.

4 years agoMinor cleanup.
Simon Forman [Mon, 27 Jan 2020 00:50:49 +0000 (16:50 -0800)]
Minor cleanup.

I feel like I should keep the un-partially-reduced thun/4 but
you can still read it, and the reduced forms are more efficient.
(And not too much more wordy.)

4 years agoExperiments with partial reduction are very promising.
Simon Forman [Sun, 26 Jan 2020 21:21:47 +0000 (13:21 -0800)]
Experiments with partial reduction are very promising.

Functions become clauses like these:

    thun(symbol(rolldown), [], [C, A, B|D], [A, B, C|D]).
    thun(symbol(rolldown), [A|B], [E, C, D|F], G) :-
        thun(A, B, [C, D, E|F], G).
    thun(symbol(dupd), [], [A, B|C], [A, B, B|C]).
    thun(symbol(dupd), [A|B], [C, D|E], F) :-
        thun(A, B, [C, D, D|E], F).
    thun(symbol(over), [], [B, A|C], [A, B, A|C]).7
    thun(symbol(over), [A|B], [D, C|E], F) :-
        thun(A, B, [C, D, C|E], F).

Definitions become

    thun(symbol(of), A, D, E) :-
        append([symbol(swap), symbol(at)], A, [B|C]),
        thun(B, C, D, E).
    thun(symbol(pam), A, D, E) :-
        append([list([symbol(i)]), symbol(map)], A, [B|C]),
        thun(B, C, D, E).
    thun(symbol(popd), A, D, E) :-
        append([list([symbol(pop)]), symbol(dip)], A, [B|C]),
        thun(B, C, D, E).

These are tail-recursive and allow for better indexing so I would expect
them to be more efficient than the originals.

Notice the difference between the original thun/4 rule for definitions
and this other one that actually works.

    thun(symbol(Def),   E, Si, So) :- def(Def, Body), append(Body, E,    E o),    thun(Eo, Si, So).
    thun(symbol(Def),   E, Si, So) :- def(Def, Body), append(Body, E, [T|Eo]), thun(T, Eo, Si, So)

The latter reduces to:

    thun(symbol(A), C, F, G) :-
        def(A, B),
        append(B, C, [D|E]),
        thun(D, E, F, G).

We want something like...

    thun(symbol(B), [], A, D) :- def(B, [H|C]), thun(H, C, A, D).
    thun(symbol(A), [H|E0], Si, So) :-
        def(A, [DH|DE]),
        append(DE, [H|E0], E),
        thun(DH, E, Si, So).

But it's good enough.  The earlier version doesn't transform into
correct code:

    thun(symbol(B), D, A, A) :- def(B, C), append(C, D, []).
    thun(symbol(A), C, F, G) :- def(A, B), append(B, C, [D|E]), thun(D, E, F, G).

It would probably be good to investigate what goes wrong there.)

It doesn't seem to work right for thun/4 combinator rules either.  Dunno what's
up there.

4 years agoPartial reduction of thun/3 in the thun/4 relation.
Simon Forman [Sun, 26 Jan 2020 20:48:38 +0000 (12:48 -0800)]
Partial reduction of thun/3 in the thun/4 relation.

It mostly works.

4 years agoMap combinator works with types.
Simon Forman [Sun, 26 Jan 2020 18:15:32 +0000 (10:15 -0800)]
Map combinator works with types.

4 years agoRemove '==' from definitions. (Bools)
Simon Forman [Sun, 26 Jan 2020 17:48:30 +0000 (09:48 -0800)]
Remove '==' from definitions.  (Bools)

I decided that the conceptual simplicity of omitting '==' is more useful
than the cosmetic value of keeping it.  The defs.txt file is now just one
definition per line with the first symbol giving the name.

Also, bools are literals.

4 years agoDefinition for 'not' in terms of 'branch'.
Simon Forman [Sun, 26 Jan 2020 16:44:57 +0000 (08:44 -0800)]
Definition for 'not' in terms of 'branch'.

4 years agoProper types, checking, inference.
Simon Forman [Sun, 26 Jan 2020 16:43:52 +0000 (08:43 -0800)]
Proper types, checking, inference.

When I first translated Joy into Prolog I was so blown away by the Prolog
unification over list structures and Triska's CLP(FD) semantics for math
(both in the sense that he wrote CLP(FD) for SWI Prolog and he suggested
it for Joy-in-Prolog specifically) that I didn't realize that it wasn't
quite type inference.

It's kind of like type inference, in that lists are handled correctly and
the CLP(FD) library creates and maintains a kind of context for
constraints, which are sort-of like "dependent types" if you squint a
little.  But you can still do things like 'cons' a number to a number and
get (a Prolog term) like [2|3] which is almost certainly a bug.

So I went through and added type "tags" as Prolog terms: int/1, list/1,
and symbol/1.  The parser assigns them and all the primitive functions
and combinators use them (and so all the definitions do too.)  With this
information Prolog can e.g. prevent attempts to add numbers and lists, and
so on.

This also allows for the thun/3 relation to be implemented a little more
efficiently (without much loss of beauty) by looking ahead one term in
the pending expression and dispatching on structural "type" in a thun/4
relation.  I miss the elegance of the previous version, but this lets
Prolog index on the structural type tags that the parser produces.

(This might mess up tail recursion because now we have a loop between
thun/3 and thun/4 but we can eliminate that problem by partial reduction
on thun/3.  TODO.)

Now that literals are tagged by the parser there's no need for literal/1.

4 years agoDon't assert defs twice.
Simon Forman [Sun, 26 Jan 2020 00:13:06 +0000 (16:13 -0800)]
Don't assert defs twice.

Each definition is getting parsed with the name of the next one as part
of its body, then the next one fails to parse and the thing backtracks.
So each definition (but the last) gets asserted twice.

    def(--,[1,-,?])
    def(--,[1,-])
    def(?,[dup,bool,++])
    def(?,[dup,bool])
    def(++,[1,+,anamorphism])
    def(++,[1,+])
    def(anamorphism,[[pop,[]],swap,[dip,swons],genrec,app1])
    def(anamorphism,[[pop,[]],swap,[dip,swons],genrec])
    def(app1,[grba,infrst,app2])
    def(app1,[grba,infrst])
    def(app2,[[grba,swap,grba,swap],dip,[infrst],cons,ii,app3])
    def(app2,[[grba,swap,grba,swap],dip,[infrst],cons,ii])

...and so on.

4 years agoChange back to CLP(FD) semantics.
Simon Forman [Sat, 25 Jan 2020 23:50:50 +0000 (15:50 -0800)]
Change back to CLP(FD) semantics.

Minor changes to parser to make it less logical but a little firmer.
(E.g. ints can't be backtracked into becoming symbols!)

4 years agoDocs and minor cleanup to the grammar.
Simon Forman [Sat, 25 Jan 2020 22:35:44 +0000 (14:35 -0800)]
Docs and minor cleanup to the grammar.

4 years agoMinor cleanup.
Simon Forman [Tue, 3 Dec 2019 16:41:42 +0000 (08:41 -0800)]
Minor cleanup.

4 years agoMake parser REs into module-level "constants".
Simon Forman [Mon, 2 Dec 2019 22:26:07 +0000 (14:26 -0800)]
Make parser REs into module-level "constants".

4 years agoSome helper scripts for windows.
Simon Forman [Mon, 2 Dec 2019 22:14:38 +0000 (14:14 -0800)]
Some helper scripts for windows.

4 years agoMinor docs update.
Simon Forman [Mon, 2 Dec 2019 22:13:50 +0000 (14:13 -0800)]
Minor docs update.

4 years agoDebugging this sucks.
Simon Forman [Thu, 28 Nov 2019 15:58:42 +0000 (07:58 -0800)]
Debugging this sucks.

Even with the RISC emu GUI.

Redesign vm?  Add more tooling?  Use MetaII?

Happy Thanksgiving!

4 years agoEmit a sort of symbol table.
Simon Forman [Thu, 14 Nov 2019 05:15:33 +0000 (21:15 -0800)]
Emit a sort of symbol table.

4 years ago4 is already an offset
Simon Forman [Wed, 13 Nov 2019 23:07:27 +0000 (15:07 -0800)]
4 is already an offset

4 years agoDefinitions.
Simon Forman [Wed, 13 Nov 2019 19:41:41 +0000 (11:41 -0800)]
Definitions.

4 years ago"swap" word.
Simon Forman [Wed, 13 Nov 2019 00:09:13 +0000 (16:09 -0800)]
"swap" word.

4 years ago"new" word.
Simon Forman [Tue, 12 Nov 2019 23:47:33 +0000 (15:47 -0800)]
"new" word.

4 years agoRefactor sub_base_merge_and_store.
Simon Forman [Tue, 12 Nov 2019 19:42:24 +0000 (11:42 -0800)]
Refactor sub_base_merge_and_store.

4 years agodup cons i
Simon Forman [Tue, 12 Nov 2019 17:23:31 +0000 (09:23 -0800)]
dup cons i

4 years agoForgot to "return" from i combinator.
Simon Forman [Tue, 12 Nov 2019 17:18:58 +0000 (09:18 -0800)]
Forgot to "return" from i combinator.

4 years agoEven "nicer".
Simon Forman [Tue, 12 Nov 2019 17:06:22 +0000 (09:06 -0800)]
Even "nicer".

4 years agoCharming.
Simon Forman [Tue, 12 Nov 2019 17:03:54 +0000 (09:03 -0800)]
Charming.

4 years agodexpr//1
Simon Forman [Tue, 12 Nov 2019 17:02:19 +0000 (09:02 -0800)]
dexpr//1

4 years agoThe i combinator.
Simon Forman [Tue, 12 Nov 2019 16:37:20 +0000 (08:37 -0800)]
The i combinator.

4 years agohead_addr
Simon Forman [Mon, 11 Nov 2019 15:51:58 +0000 (07:51 -0800)]
head_addr

4 years agoSimple push of empty list.
Simon Forman [Mon, 11 Nov 2019 15:45:22 +0000 (07:45 -0800)]
Simple push of empty list.

4 years agoDup.
Simon Forman [Sun, 10 Nov 2019 22:35:57 +0000 (14:35 -0800)]
Dup.

ANd portray_clause to stablize output logical variable names.

4 years agohalt.
Simon Forman [Sun, 10 Nov 2019 18:58:47 +0000 (10:58 -0800)]
halt.

4 years agoword works with negative numbers now.
Simon Forman [Sun, 10 Nov 2019 18:53:55 +0000 (10:53 -0800)]
word works with negative numbers now.

symbols moved to head of machine code.

4 years agomerge_and_store, chain_link
Simon Forman [Sun, 10 Nov 2019 05:55:01 +0000 (21:55 -0800)]
merge_and_store, chain_link

4 years agoMinor refactor.
Simon Forman [Sun, 10 Nov 2019 05:20:31 +0000 (21:20 -0800)]
Minor refactor.

It doesn't save space (but if I reuse it once it will.)

4 years agoPass through label.
Simon Forman [Sun, 10 Nov 2019 04:19:15 +0000 (20:19 -0800)]
Pass through label.

4 years agoif_literal and lookup
Simon Forman [Sun, 10 Nov 2019 02:39:39 +0000 (18:39 -0800)]
if_literal and lookup

4 years agoincr stack
Simon Forman [Sun, 10 Nov 2019 02:28:01 +0000 (18:28 -0800)]
incr stack

Really decr, but I'm abstracting.

4 years agoRefactoring and cleanup.
Simon Forman [Sun, 10 Nov 2019 02:15:56 +0000 (18:15 -0800)]
Refactoring and cleanup.

4 years agoload
Simon Forman [Sun, 10 Nov 2019 02:02:07 +0000 (18:02 -0800)]
load

4 years agoBleah.
Simon Forman [Sat, 9 Nov 2019 23:20:06 +0000 (15:20 -0800)]
Bleah.

4 years agoBase address in unpack_pair.
Simon Forman [Sat, 9 Nov 2019 23:12:44 +0000 (15:12 -0800)]
Base address in unpack_pair.

4 years agounpack_pair
Simon Forman [Sat, 9 Nov 2019 23:03:17 +0000 (15:03 -0800)]
unpack_pair

4 years agoMinor refactor.
Simon Forman [Sat, 9 Nov 2019 22:02:18 +0000 (14:02 -0800)]
Minor refactor.

4 years agoRefactoring, with oddball quoting "symbols".
Simon Forman [Sat, 9 Nov 2019 21:59:06 +0000 (13:59 -0800)]
Refactoring, with oddball quoting "symbols".

4 years agoThat works nicely, again.
Simon Forman [Sat, 9 Nov 2019 21:42:14 +0000 (13:42 -0800)]
That works nicely, again.

4 years agoConvert to ? DCG and it's macro-time!
Simon Forman [Sat, 9 Nov 2019 21:34:51 +0000 (13:34 -0800)]
Convert to ? DCG and it's macro-time!

4 years agoI think that does it for cons.
Simon Forman [Sat, 9 Nov 2019 20:14:51 +0000 (12:14 -0800)]
I think that does it for cons.

Offsets in pair records can be negative.

4 years agoThat's the mainloop converted to permit negative offsets.
Simon Forman [Sat, 9 Nov 2019 20:03:13 +0000 (12:03 -0800)]
That's the mainloop converted to permit negative offsets.

4 years agoMinor cleanup, bug fixes.
Simon Forman [Sat, 9 Nov 2019 19:27:29 +0000 (11:27 -0800)]
Minor cleanup, bug fixes.

4 years agoMinor bugfix.
Simon Forman [Fri, 8 Nov 2019 23:24:09 +0000 (15:24 -0800)]
Minor bugfix.

asr not ror.

4 years agoCons
Simon Forman [Fri, 8 Nov 2019 22:06:28 +0000 (14:06 -0800)]
Cons

4 years agoJust do it in asm.
Simon Forman [Fri, 8 Nov 2019 16:08:53 +0000 (08:08 -0800)]
Just do it in asm.

4 years agohmm...
Simon Forman [Fri, 8 Nov 2019 04:21:26 +0000 (20:21 -0800)]
hmm...

4 years agoWIth push2 finished that's the mainloop converted.
Simon Forman [Fri, 8 Nov 2019 01:25:04 +0000 (17:25 -0800)]
WIth push2 finished that's the mainloop converted.

4 years agolookup
Simon Forman [Fri, 8 Nov 2019 00:28:28 +0000 (16:28 -0800)]
lookup

4 years agoif_literal
Simon Forman [Fri, 8 Nov 2019 00:15:41 +0000 (16:15 -0800)]
if_literal

4 years agoA start on converting the mainloop.
Simon Forman [Thu, 7 Nov 2019 23:53:07 +0000 (15:53 -0800)]
A start on converting the mainloop.

4 years agoTake two on the compiler.
Simon Forman [Thu, 7 Nov 2019 22:57:41 +0000 (14:57 -0800)]
Take two on the compiler.

4 years agoMinor cleanup.
Simon Forman [Thu, 7 Nov 2019 15:55:01 +0000 (07:55 -0800)]
Minor cleanup.

4 years agoModify error reporting a lil; words word.
Simon Forman [Thu, 7 Nov 2019 15:54:04 +0000 (07:54 -0800)]
Modify error reporting a lil; words word.