* sort.c (closureless): Use scm_eval_body.

[bpt/guile.git] / libguile / ChangeLog

1999-08-26  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	This change extends the representation of primitive procedures
	with more data fields, e.g. a place for documentation and true
	procedure properties.
	
	* procs.c, procs.h (scm_subr_entry): New type: Stores data
	associated with subrs.
	(SCM_SUBRNUM, SCM_SUBR_ENTRY, SCM_SUBR_GENERIC, SCM_SUBR_PROPS,
	SCM_SUBR_DOC): New macros.
	(scm_subr_table): New variable.
	(scm_mark_subr_table): New function.
	
	* init.c (scm_boot_guile_1): Call scm_init_subr_table.

	* gc.c (scm_gc_mark): Don't mark subr names here.
	(scm_igc): Call scm_mark_subr_table.


	This change implements a scheme for letting a generic work as a
	shadow for a primitive procedure.  If the primitive procedure
	can't dispatch on its arguments, control is left over to the
	generic.  Normal wrong type arg errors will be generated until the
	user has hung the first method on the primitive.

	* snarf.h (SCM_GPROC, SCM_GPROC1): New macros.

	* procs.c, procs.h (scm_subr_p): New function (used internally).

	* gsubr.c, gsubr.h (scm_make_gsubr_with_generic): New function.

	* objects.c, objects.h (scm_primitive_generic): New class.

	* objects.h (SCM_CMETHOD_CODE, SCM_CMETHOD_ENV): New macros.

	* print.c (scm_iprin1): Print primitive-generics.

	* __scm.h (SCM_WTA_DISPATCH_1, SCM_GASSERT1,
	SCM_WTA_DISPATCH_2, SCM_GASSERT2): New macros.
	
	* eval.c (SCM_CEVAL, SCM_APPLY): Replace scm_wta -->
	SCM_WTA_DISPATCH_1 for scm_cxr's (unary floating point
	primitives).  NOTE: This means that it is now *required* to use
	SCM_GPROC1 when creating float scm_cxr's (float scm_cxr's is an
	obscured representation that will be removed in the future anyway,
	so backward compatibility is no problem here).

	* numbers.c: Converted most numeric primitives (all but bit
	comparison operations and bit operations) to dispatch on generic
	if args don't match.

	
	Better support for applying generic functions.

	* eval.c, eval.h (scm_eval_body): New function.

	* objects.c (scm_call_generic_0, scm_call_generic_1,
	scm_call_generic_2, scm_call_generic_3, scm_apply_generic): New
	functions.
	
	
	Optimization of the generic function dispatch mechanism.
	
	* eval.c (SCM_CEVAL): Apply the cmethod directly after having
	called scm_memoize_method instead of doing a second lookup.

	* objects.h (scm_memoize_method): Now returns the memoized cmethod.


	Bugfix

	* procs.c (scm_make_subr_opt): Use scm_sysintern0 instead of
	scm_sysintern so that the binding connected with the subr name
	isn't cleared when we give set = 0.

	
1999-08-24  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	More transparent handling of ports with print states.
	
	* print.h (SCM_PORT_WITH_PS_P, SCM_PORT_WITH_PS_PORT,
	SCM_PORT_WITH_PS_PS): Represent ports with print states as a smob
	instead of a pair of a port and a print state.  We'll need to cons
	once extra in scm_printer_apply but the type system will be
	cleaner, it will mix better with GOOPS, and, it will be even more
	transparent to the user.

	* print.c (scm_get_print_state): New procedure: Given an output
	port, return the print state associated to it in the current print
	chain, if one exists;
	(scm_port_with_print_state): New procedure: Associate a
	print-state with a port.
	(scm_valid_oport_value_p): Use SCM_PORT_WITH_PS_P;
	(scm_printer_apply): Wrap port and pstate as a smob;
	(print_state_printer): Removed.
	
	* objects.c (scm_class_of): Treat scm_tc16_port_with_ps as ports.

	* eval.c (scm_init_eval): Use scm_make_smob_type instead of
	scm_newsmob.

	* ports.c (scm_output_port_p): Bugfix: Coerce output port before
	testing (otherwise the port-print-state trick won't be transparent
	to the user; one example where this caused problems was in the
	(ice-9 format) module).

1999-08-23  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (SCM_CEVAL): Let the SCM_IM_SLOT_SET_X form return
	SCM_UNSPECIFIED instead of the set value.

1999-08-20  James Blandy  <jimb@mule.m17n.org>

	* load.c (scm_init_load_path): Remove support for SCHEME_LOAD_PATH.

	* ports.h (enum scm_port_rw_active): New enum, containing
	SCM_PORT_READ, SCM_PORT_WRITE, and SCM_PORT_NEITHER (instead of
	zero).  The debugger knows about enums, but doesn't know about
	#defines.
	(typedef scm_port): Declare rw_active member to be an enum
	scm_port_rw_active.
	* fports.c (fport_flush, fport_end_input): Use SCM_PORT_NEITHER
	instead of zero.
	* ports.c (scm_add_to_port_table): Same.
	* strports.c (st_flush, st_end_input): Same.

	* ioext.c (scm_do_read_line, scm_read_line): Use scm_must_malloc,
	scm_must_realloc, and scm_done_malloc as appropriate. 

1999-08-20  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* sort.c (quicksort): Added condition to protect the algorithm
	from crashing the interpreter if the less predicate is buggy.

1999-08-19  Gary Houston  <ghouston@easynet.co.uk>

	* fports.c (fport_write): fix line-buffering mode again.
	(scm_open_file): recognise 'l' for line-buffering.
	(scm_setvbuf): recognise _IOLBF for line-buffering.

1999-08-19  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* Makefile.am (libguile_la_LDFLAGS): Increased the version number
	of libguile to 5.0.

	* eval.c (SCM_APPLY), sort.c (closureless): Expand body when
	evaluating closures.

1999-08-18  Gary Houston  <ghouston@easynet.co.uk>

	* fports.c (fport_write): use memcpy instead of strncpy, in case
	the data contains NUL.

1999-08-17  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* gh.h (gh_vector_to_list): Bugfix.  (Thanks to Frank Cieslok.)
	
	* backtrace.c, debug.c, eval.c, eval.h, gsubr.c, procprop.h,
	read.c, srcprop.c, srcprop.h (scm_i_filename, scm_i_line,
	scm_i_column, scm_i_copy, scm_i_name, scm_i_lambda, scm_i_source,
	scm_i_more, scm_i_procname, scm_i_dot, scm_i_arrow, scm_i_else,
	scm_i_unquote, scm_i_uq_splicing, scm_i_apply, scm_i_enter_frame,
	scm_i_apply_frame, scm_i_exit_frame, scm_i_trace, scm_i_quote,
	scm_i_begin, scm_i_if, scm_i_and, scm_i_or, scm_i_case,
	scm_i_cond, scm_i_letstar, scm_i_do, scm_i_quasiquote,
	scm_i_define, scm_i_letrec, scm_i_let, scm_i_atapply,
	scm_i_atcall_cc, scm_i_breakpoint): Renamed: Consequently use
	scm_sym_ as prefix for symbols.

	* debug.c (scm_i_proc, scm_i_args, scm_i_eval_args): Removed.

	* eval.c, eval.h (scm_sym_begin, scm_sym_if, scm_sym_and,
	scm_sym_case, scm_sym_cond, scm_sym_letstar, scm_sym_do,
	scm_sym_define, scm_sym_letrec, scm_sym_atapply,
	scm_sym_atcall_cc): Made global.
		
1999-08-16  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (scm_sym_args): Made global.

	* objects.c (scm_set_object_procedure_x): Disallow setting of
	procedures for pure generic functions.

1999-08-12  Gary Houston  <ghouston@easynet.co.uk>

	* ports.c (scm_seek): one more: was scm_lseek.  Also changed the
	Scheme name from lseek to seek, but lseek was added recently so
	it shouldn't be a big problem.
	* ports.c, gdbint.c, ioext.c: changed callers.

1999-08-11  Gary Houston  <ghouston@easynet.co.uk>

	* fports.c (fport_input_waiting): if select is used, return 1
	instead of whatever FD_ISSET expands to.  maybe it will be useful
	to interpret the value from the input_waiting ptob procedure as a
	lower bound on the number of bytes available.

	* Mikael asked for a few names to be changed...

	* ports.c (scm_make_port_type): take the write procedure as the 
	second argument instead of the flush procedure.
	* ports.h (scm_ptob_descriptor): rename the ptob procedures:
	fflush -> flush,  read_flush -> end_input,  fclose -> close,
	fill_buffer -> fill_input,  ftruncate -> truncate,
	input_waiting_p -> input_waiting.

	* ports.c (end_input_void_port): was read_flush_void_port.
	(scm_set_port_end_input): was scm_set_port_flush_input.
	(scm_set_port_flush): was scm_set_port_write.
	(scm_set_port_input_waiting): was scm_set_port_input_waiting_p
	(scm_end_input): was scm_read_flush.
	(scm_fill_input): was scm_fill_buffer.
	(scm_flush): was scm_fflush.
	* fports.c (fport_input_waiting): renamed from fport_input_waiting_p.
	(fport_end_input): was local_read_flush.
	(fport_flush): was local_fflush.
	(fport_close): was local_fclose.
	(fport_truncate): was local_ftruncate.
	(fport_seek): was local_seek.
	(fport_free): was local_free.
	(fport_fill_input): was fport_fill_buffer.
	* strports.c (st_end_input): was st_read_flush.
	(st_truncate): was st_ftruncate.
	* vports.c: (sf_flush): was sfflush.
	(sf_close): was sfclose.
	(sf_fill_input): was sf_fill_buffer.
	
	* ports.c, fports.c, strports, vports.c, ioext.c, unif.c, filesys.c:
	change callers.

1999-08-06  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (SCM_IM_DISPATCH): Rewrote dispatch protocol.  Dispatch
	forms now contain the expressions to be dispatched upon instead of
	depending on a surrounding lambda or let; Generic function
	dispatch has been optimized; `apply' on a generic function now
	works a little bit strangely.  It uses a trick so that the type
	dispatch code in SCM_CEVAL can be reused.

	* objects.h, objects.c (scm_apply_generic_env): Added (used by
	apply).
	(scm_operator_p): Added.
	(scm_sym_atdispatch): Added.
	(scm_set_object_procedure_x): Modified to handle the new style
	generic functions.
	(scm_object_procedures): New debugging procedure.

1999-08-05  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c, eval.h (scm_sym_args): Added.

	* objects.h (SCM_CLASSF_PURE_GENERIC): Added.

	* feature.c, feature.h (scm_c_run_hook): Added.

	* eval.c (SCM_CEVAL:SCM_IM_DISPATCH): Bugfix: Jump back to
	cdrxnoap and loopnoap instead of begin and loop.

1999-08-04  Gary Houston  <ghouston@easynet.co.uk>

	* ports.c (scm_putc, scm_puts),
	* unif.c (scm_uniform_array_write): use scm_lfwrite.
	* ports.c (scm_putc): change type of first argument from int to	char.
	
1999-08-04  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (SCM_CEVAL): Improvements to SCM_IM_DISPATCH and
	SCM_IM_HASH_DISPATCH.

	* objects.h (SCM_CLASSF_GOOPS_VALID): Added.
	(scm_si_redfined, scm_si_hashsets): Moved.

	* objects.c (scm_class_of): Use the new SCM_CLASSF_GOOPS_VALID
	flag which combines type and status info so that the class
	redefinition protocol has zero cost during normal execution.

1999-08-03  Gary Houston  <ghouston@easynet.co.uk>

	* ports.h (scm_ptob_descriptor): include a write procedure again.
	it's more efficient for unbuffered fports (e.g., sockets.)

	* ports.c (scm_puts): use ptob->write.
	* vports.c (scm_make_sfptob): set write proc in ptob.
	* strports.c (scm_make_stptob): set write proc in ptob.
	* ports.c (write_void_port): new procedure.
	* vports.c (sf_write): new procedure.
	* ports.c (scm_lfwrite): use ptob->write.
	* strports.c (st_write): new procedure.
	* fports.c (fport_write): new procedure.
	(scm_make_fptob): set write in ptob to fport_write.
	* ports.h: prototype for scm_set_port_write.
	* ports.c (scm_make_port_type): initialise ptob write procedure.
	(scm_set_port_write): new proc.

1999-08-01  Jim Blandy  <jimb@savonarola.red-bean.com>

	* ports.c (scm_char_ready_p): Don't try to find PORT's ptab entry
	until we've verified that it is actually a port.  (Thanks to
	Lorentey Karoly.)

1999-07-31  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* gc.c (scm_must_malloc, scm_must_realloc): Removed unnecessary
	code, particularly an unnecessary test (len != size, where len ==
	size).  (Was this leftovers from debugging code, or have I missed
	something profound?)

	* hashtab.c: Bugfix: Don't declare s_hash_fold without storage
	size. (Thanks to James Dean Palmer.)

	* numbers.c (scm_makdbl): Bugfix: Initialize imaginary part.
	(Thanks to Lorentey Karoly.)

1999-07-30  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (scm_m_expand_body): Use scm_cons_source.

	* struct.c (scm_print_struct): Use vtable name.

	* print.c (scm_init_print): Set name of print state type.

	* stacks.c (scm_init_stacks): Set name of stack type.

1999-07-29  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* eval.c (SCM_CEVAL): Removed old implementation of internal
	define.

	* gsubr.c, procprop.h (scm_i_inner_name): Removed.

	* debug.c, debug.h (scm_reverse_lookup): Added.
	(scm_procedure_name): Use scm_reverse_lookup to lookup the name of
	internal procedure definitions; Don't use scm_i_inner_name.

	* eval.c, tags.h, print.c (SCM_IM_SLOT_REF, SCM_IM_SLOT_SET_X):
	New isym operations.

	* eval.h: Added prototypes for multi language support functions.

	* eval.c (SCM_IM_DISPATCH, SCM_IM_HASH_DISPATCH): Don't use
	improper lists in the low-level representation, since that will
	cause a begin to be prepended at macro expansion.

	* eval.c (scm_cons_source): Version of cons which copies source
	properties from an existing cell.
	(scm_copy_tree, SCM_CEVAL): Use	scm_cons_source.

	* debug.c (scm_procedure_source): Cons SCM_IM_LAMBDA onto
	procedure source before calling scm_unmemocopy instead of faking
	an environment.

1998-10-25  Marius Vollmer  <mvo@zagadka.ping.de>

	Ported `internal defines' fix from SCM.  Original ChangeLog entry:

	    1998-07-09  Radey Shouman  <radey@colorage.com>

	    * eval.c (ceval_1): Modifications to allow rewriting of interal
	    DEFINE to LETREC: If an ISYM is evaluated in non-tail position the
	    body of which it is the CAR is macro expanded by m_expand_body,
	    which rewrites internal DEFINE.

	    (m_expand_body): Added.

	    (m_macroexp1): Added argument to control error checking:
	    m_expand_body may speculatively expand forms in the wrong
	    environments.  Made argument number checks conditional on
	    RECKLESS.

	    (m_body): Added, error checks bodies and inserts the ISYM tokens.

	    (m_lambda): (m_letstar): (m_letrec1): (m_letrec): (m_let): Now
	    call m_body.

	    (m_cond): (m_case): (m_quote): Modified to avoid destructively
	    changing their argument forms.  Since m_expand_body
	    speculatively macro expands forms the process must be
	    reversible.

	    (m_ident_eqp): Fixed to use proper environment.

	    (renamed_ident): Added DEFER_INTS_EGC.

	    Added prototypes for static functions.

	* eval.c

	(undef_cell): New.
	
	(scm_lookupcar1, scm_lookupcar): Added CHECK argument.  When CHECK
	is false, do not produce an error for unbound variables, return a
	pointer to cell_undef instead.

	(EVALCELLCAR, XEVALCAR): Call scm_lookupcar with check=1.

	(scm_m_body): New.

	(scm_m_cond, scm_m_case, scm_m_quote): Modified to avoid
	destructively changing their argument forms.  Since m_expand_body
	speculatively macro expands forms the process must be reversible.

	(scm_m_lambda): Use scm_m_body instead of bodycheck.  Account for
	SCM_IM_LET introduced by named lets.

	(scm_m_letstar): Use scm_m_body instead of bodycheck.

	(scm_m_letrec1, scm_letrec): Split scm_letrec into scm_letrec1 and
	scm_letrec.  scm_letrec1 does not check for a null binding and
	takes an additional argument to specify the ISYM of the body.  Use
	scm_m_body instead of bodycheck.

	(scm_m_let): Use scm_m_body instead of bodycheck.

	(scm_m_expand_body, scm_macroexp): New.

	(unmemocopy): Account for ISYMs introduced by scm_m_body.

	(ceval, deval): Call scm_m_expand_body.  Call scm_lookupcar with
	check=1.  Throw error for internal defined that have not been
	rewritten by scm_m_expand_body.

	* eval.h: Added prototypes for scm_m_expand_body and scm_macroexp.
	Removed prototype for SCM_APPLY.

	* tags.h: Added extern declaration of scm_isymnames.

1999-07-27  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* Makefile.am (EXTRA_libguile_la_SOURCES): Added lang.c.

	* lang.c: New file: Beginning of multi-language support.
	
	* init.c (scm_boot_guile_1): Added call to scm_init_lang ().

	* dynwind.c (scm_dowinds): Removed obsolete wind_key #f case.
	(scm_dynamic_wind): Added argument checking for the after guard so
	that we don't add garbage on the dynwind chain.
	(scm_swap_bindings): Added.

	* tags.h, print.c (SCM_IM_NIL_COND, SCM_IM_NIL_IFY, SCM_IM_T_IFY,
	SCM_IM_0_COND, SCM_IM_0_IFY, SCM_IM_1_IFY), print.c
	(scm_isymnames): New isyms for multi-language support.
	
 	* eval.c (scm_nil, scm_t): New symbols.
	(nil-cond, nil-ify, t-ify, 0-cond, 0-ify, 1-ify): New special
	forms for multi-language support.

1999-07-25  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* random.c, random.h (scm_c_default_rstate, scm_c_uniform32):
	Added.
	Renamed functions in the random function library interface
	from scm_i_XXX --> scm_c_XXX.

1999-07-25  Gary Houston  <ghouston@easynet.co.uk>

	* ports.c (scm_putc): fix line-buffering.

1999-07-25  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* ports.c, ports.h, fports.c, strports.c, vports.c: Renamed
	scm_set_ptob_XXX --> scm_set_port_XXX.

1999-07-24  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* ports.c, ports.h (scm_make_port_type): New interface for
	creation of port types (replaces scm_newptob).  Just as for the
	smobs, we need to separate the internal representation of smob
	types from the interface, so that we easily can add new fields and
	rearrange things without caring about backward compatibility.
	This change was forced by the need in GOOPS to create classes
	representing port types.
	(scm_set_ptob_mark, scm_set_ptob_free, scm_set_ptob_print,
	scm_set_ptob_equalp, scm_set_ptob_flush_input, scm_set_ptob_close,
	scm_set_ptob_seek, scm_set_ptob_truncate,
	scm_set_ptob_input_waiting_p): New setters.
	(scm_newptob): Rewritten to use scm_make_port_type.  For backward
	compatibility.
	(scm_ptobs): Changed type scm_ptobfuns --> scm_ptob_descriptor.
	(scm_prinport): Removed.
	(scm_port_print): Added.
	(scm_print_port_mode): Added.
	(void_port_ptob, print_void_port, close_void_port, noop0):
	Removed.  Removed #include "genio.h" Added #include "objects.h",
	#include "smobs.h"

	* fports.c (prinfport): Moved code from ports.c.
	(local_free): Added.
	(scm_fptob): Removed.  Instead use new interface.
	(scm_make_fptob): Added.  (Need to create basic ports in a
	specific order in ports.c.)
	
	* strports.c (scm_stptob, prinstpt, noop0): Removed
	(scm_make_stptob): Added.

	* vports.c (scm_sfport, prinsfpt, sf_read_flush, noop0): Removed.
	(scm_make_sfport): Added.
	
	* filesys.c (scm_dir_print): Don't use the port printing code.
	Instead provide specific directory printer.
	
	* gc.c (scm_gc_sweep): Use value returned from scm_ptobs[].free.

	* ioext.c (scm_redirect_port): Replaced scm_ptobfuns -->
	scm_ptob_descriptor.

	* smob.c (scm_smob_print): Handle non-existing type name nicely.
	Removed #include "genio.h"

	* objects.c (scm_make_port_classes): New function ptr.

1999-07-24  Gary Houston  <ghouston@easynet.co.uk>

	* gdbint.c (gdb_print, gdb_read): call scm_truncate_file.

	* ports.c (scm_truncate_file): renamed from scm_ftruncate.
	allow the 1st argument to be a fdes or filename as well as a
	port (as in the filesys.c version).

	* filesys.c (scm_truncate_file): removed.

1999-07-24  Mikael Djurfeldt  <mdj@thalamus.nada.kth.se>

	* readline.c, readline.h: Removed.

	* objects.c, objects.h (scm_port_class): Added.
	(scm_class_of): Look up port class in scm_port_class.
	(SCM_IN_PCLASS_INDEX, SCM_OUT_PCLASS_INDEX,
	SCM_INOUT_PCLASS_INDEX): Added.

	
	* Makefile.am: Removed genio.c, genio.x.

	* genio.c: Removed.

1999-07-23  Mikael Djurfeldt  <mdj@mdj-pc.nada.kth.se>

	* init.c: Make sure that scm_post_boot_init_modules is called only
	once.  (Important when using a dumped image.; Thanks to Bernard
	Urban.)

1999-07-19  Jim Blandy  <jimb@savonarola.red-bean.com>

	* guardians.c (scm_guardian_zombify): Separate scanning for
	zombies from marking the pairs of the free list.
	
	* guardians.c (scm_guardian_zombify): Don't set marks manually ---
	use the macros.  (Thanks to Michael Livshin.)

	* eval.c (scm_m_lambda): Let bodycheck check the body of the
	lambda.  Let your sins be purified by the blood of the lambda.
	(Thanks to Eric Hanchrow.)

	* net_db.c (h_errno): Don't declare this if it's #defined.  Eew.
	(Thanks to Valdis Kletnieks.)