1		% (c) 2020-2024 Lehrstuhl fuer Softwaretechnik und Programmiersprachen,
2		% Heinrich Heine Universitaet Duesseldorf
3		% This software is licenced under EPL 1.0 (http://www.eclipse.org/org/documents/epl-v10.html)
4
5		:- module(tools_fastread, [ read_term_from_stream/2,
6		read_term_from_string/2,
7		read_test/1, tok_test/1,
8		fastrw_read/3
9		]).
10
11		:- use_module(module_information,[module_info/2]).
12
13		:- module_info(group,infrastructure).
14		:- module_info(description,'This module provides a more efficient way to parse and read large Prolog terms.').
15
16		:- use_module(self_check).
17		:- use_module(error_manager,[add_error/3]).
18
19
20		:- use_module(error_manager).
21
22
23		:- use_module(library(fastrw),[fast_read/2]).
24		% this reads terms in binary format
25		fastrw_read(S,Term,Error) :-
26		catch( fast_read(S,Term1), % from library fastrw
27		error(E,_),
28		( E=syntax_error(_) -> Error = true
29		; E=permission_error(_,_,_) -> Term1 = end_of_file
30		; E=consistency_error(_,_,_)
31		-> add_error(tools_fastread,'Consistency error when reading from stream: ',E),
32		Error = true
33		;
34		add_error(tools_fastread,'Unknown error when reading from stream: ',E),
35		throw(E))
36		),
37		Term1 = Term.
38
39
40
41
42		:- set_prolog_flag(double_quotes, codes).
43
44		% Limitations:
45		% - only prefix notation and lists supported
46		% - supports atoms with and without quotes, in quotes one can use octal and hex literals and backslash escaping
47		% - no support for variables (yet?)
48		% - no support for operator declarations
49		% - should work with numbers (integers and floats)
50		% - no real support for (a,b) terms using , as infix
51		% - no support for block comments
52
53
54		% --------------------
55
56		% Tokeniser
57
58
59		:- use_module(tools,[print_memory_used_wo_gc/0]).
60		tok_read(File) :- statistics(walltime,_),
61		%open(File,read,Stream1),reset_line_nr, repeat, get_code(Stream1,C), C = -1,!, close(Stream1),
62		%statistics(walltime,[_,W1]), format('~n% Walltime ~w ms to get_codes ~w~n',[W1,File]),
63		open(File,read,Stream),reset_line_nr,
64		read_tokens(Stream),close(Stream),
65		statistics(walltime,[_,W2]), format('~n% Walltime ~w ms to tokenise ~w~n',[W2,File]),
66		print_memory_used_wo_gc.
67
68		% set_prolog_flag(profiling,on)
69		tok_test(1) :-
70		tok_read('/Users/leuschel/git_root/prob_examples/public_examples/B/Demo/Bakery0.prob').
71		tok_test(2) :-
72		tok_read('/Users/leuschel/git_root/prob_examples/public_examples/B/Benchmarks/CAN_BUS_tlc.prob').
73		tok_test(3) :-
74		tok_read('/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated4000.prob').
75		tok_test(4) :-
76		tok_read('/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.prob').
77		tok_test(7) :-
78		tok_read('/Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.prob').
79
80		read_tokens(Stream) :-
81		get_next_token(Stream,Tok), % print_token(Tok),
82		!,
83		(Tok = eof -> true
84		; read_tokens(Stream)).
85		read_tokens(Stream) :-
86		get_code(Stream,C),
87		format('Could not process: ~w~n',[C]).
88
89		%print_token(atom(L)) :- !, format('Token: Atom ~s~n',[L]).
90		%print_token(S) :- !, format('Token: ~w~n',[S]).
91
92		% get next Prolog token from a stream
93		% TO DO: produce a version without peek, which returns first code following token
94		get_next_token(Stream,Token) :- get_code(Stream,C), !,
95		get_next_token1(C,Stream,Token).
96
97
98		get_next_token1(-1,_,eof).
99		get_next_token1(32,Stream,Token) :- !, get_next_token(Stream,Token).
100		get_next_token1(10,Stream,Token) :- !, inc_line_nr, get_next_token(Stream,Token).
101		get_next_token1(13,Stream,Token) :- !, get_next_token(Stream,Token).
102		get_next_token1(0'(,_,open_paren). % adding a cut here seems to degrade performance, but leads to backtracking
103		get_next_token1(0'),_,close_paren).
104		get_next_token1(0'[,_,open_bracket).
105		get_next_token1(0'],_,close_bracket).
106		get_next_token1(0',,_,comma).
107		get_next_token1(0'.,_,dot).
108		get_next_token1(39,Stream,atom(T)) :- !, get_quoted_atom_codes(Stream,T).
109		get_next_token1(0'%,Stream,Token) :- !, % line comment
110		skip_until_end_of_line(Stream),!,get_next_token(Stream,Token).
111		get_next_token1(X,Stream,number([X|T])) :- is_digit(X),!, get_number_codes(Stream,T).
112		get_next_token1(X,Stream,atom([X|T])) :- is_lowcase(X),!, get_atom_codes(Stream,T).
113		get_next_token1(X,_,_) :- prolog_code_error(X),fail.
114
115		get_number_codes(Stream,[H|T]) :- peek_code(Stream,H), is_digit(H),
116		!, get_code(Stream,_), get_number_codes(Stream,T).
117		get_number_codes(_,[]).
118		% TO DO: also accept floats
119
120		get_atom_codes(Stream,[H|T]) :-
121		peek_code(Stream,H), is_unquoted_atom_letter(H),!,
122		get_code(Stream,_), get_atom_codes(Stream,T).
123		get_atom_codes(_,[]).
124
125		get_quoted_atom_codes(Stream,Res) :- get_code(Stream,H),
126		( H = 92 -> get_code(Stream,H2), % backslash: read next code and treat it specially
127		treat_escaped_code(H2,Stream,Res)
128		; H = 39 -> Res=[]
129		; H = -1 -> format('Unterminated quoted atom: ~w~n',[H]),fail
130		; Res=[H|T], get_quoted_atom_codes(Stream,T)).
131
132		skip_until_end_of_line(Stream) :- get_code(Stream,H),
133		(H = 10 -> inc_line_nr
134		; H = -1 -> format('Unterminated % comment~n',[]), prolog_token_error(H), fail
135		; skip_until_end_of_line(Stream)).
136
137		% see page 60, Section 4.1.7.6 Escape Sequences in SICStus Prolog manual
138		treat_escaped_code(Code,Stream,[NewCode|T]) :- escape_conversion(Code,NewCode),!,
139		get_quoted_atom_codes(Stream,T).
140		treat_escaped_code(Octal,Stream,[OctResult|T]) :- is_octal_digit(Octal,Nr),!, % treat chars like \374\
141		get_code(Stream,NxtCode),
142		get_octal_codes(NxtCode,Stream,Nr,OctResult),
143		get_quoted_atom_codes(Stream,T).
144		treat_escaped_code(0'x,Stream,[HexResult|T]) :- !, % treat chars like \x21D4\
145		get_code(Stream,NxtCode),
146		get_hex_codes(NxtCode,Stream,0,HexResult),
147		get_quoted_atom_codes(Stream,T).
148		treat_escaped_code(10,Stream,T) :- !, %\LFD ignored ; TO DO: also support 13
149		get_quoted_atom_codes(Stream,T).
150		treat_escaped_code(-1,_,_) :- !, format('Unterminated quoted atom ending with backslash at: ~w~n',[end_of_file]),fail.
151		treat_escaped_code(X,_,_) :- prolog_code_error('escape sequence',X),fail.
152
153		escape_conversion(39,39). % single quote '
154		escape_conversion(34,34). % double quote "
155		escape_conversion(92,92).
156		escape_conversion(0'n,10).
157		escape_conversion(0't,9).
158		escape_conversion(0'b,8).
159		escape_conversion(0'r,13).
160
161
162		get_octal_codes(H,Stream,Acc,Res) :- is_octal_digit(H,HNr),
163		!, A2 is Acc*8+HNr,
164		get_code(Stream,NxtCode), get_octal_codes(NxtCode,Stream,A2,Res).
165		get_octal_codes(92,_,Acc,Res) :- !, Res=Acc.
166		get_octal_codes(X,_,_,_) :- prolog_code_error('octal literal',X),fail.
167
168		get_hex_codes(H,Stream,Acc,Res) :- is_hex_digit(H,HNr),
169		!, A2 is Acc*16+HNr,
170		get_code(Stream,NxtCode), get_hex_codes(NxtCode,Stream,A2,Res).
171		get_hex_codes(92,_,Acc,Res) :- !, Res=Acc.
172		get_hex_codes(X,_,_,_) :- prolog_code_error('hex literal',X),fail.
173
174
175
176		% 45 = leading minus sign
177		is_digit(X) :- X=45 ; (X>=48, X=<57).
178
179		:- assert_must_succeed(tools_fastread:is_octal_digit(0'7,7)).
180		is_octal_digit(X,Nr) :- X>=48, X=<55, Nr is X-48.
181		:- assert_must_succeed(tools_fastread:is_hex_digit(0'f,15)).
182		:- assert_must_succeed(tools_fastread:is_hex_digit(0'F,15)).
183		:- assert_must_succeed(tools_fastread:is_hex_digit(0'9,9)).
184		is_hex_digit(X,Nr) :- X>=48, X=<57, Nr is X-48.
185		is_hex_digit(X,Nr) :- X>=65, X=<70, Nr is X-55. % upper-case A-F
186		is_hex_digit(X,Nr) :- X>=97, X=<102, Nr is X-87. % lower-case a-f
187		is_lowcase(X) :- X>=97, X=<122.
188
189		is_unquoted_atom_letter(X) :- X>=97, X=<122.
190		is_unquoted_atom_letter(X) :- X>=65, X=<90.
191		is_unquoted_atom_letter(X) :- X=95.
192		is_unquoted_atom_letter(X) :- X>=48, X=<57.
193
194		%is_quoted_atom_letter(X) :- X \= 39. %, X \=10, X \=13.
195
196		% skip over whitespace and return code after whitspace (peeked)
197		skip_ws(Stream,NextCode) :- peek_code(Stream,C),
198		(is_skip_whitespace(C) -> get_code(Stream,_),skip_ws(Stream,NextCode)
199		; NextCode=C).
200
201		is_skip_whitespace(32).
202		is_skip_whitespace(10) :- inc_line_nr.
203		is_skip_whitespace(13).
204
205		% ------------------------
206
207		expect_token(Stream,ETok) :- get_next_token(Stream,RTok),!,
208		(RTok = ETok -> true
209		; cur_line(L),
210		format(user_error,'Unexpected token on line ~w: ~w, expected ~w~n',[L,RTok,ETok]),
211		fail).
212
213		prolog_token_error(Tok) :- cur_line(L),
214		%format(user_error,'Unexpected token on line ~w: ~w~n',[L,Tok]),
215		add_error(fastread,'Unexpected token on line:',Tok:L).
216		prolog_code_error(Code) :- cur_line(L),
217		format(user_error,'Unexpected character on line ~w: ~w "~s"~n',[L,Code,[Code]]),
218		add_error(fastread,'Unexpected character code on line:',Code:line(L)).
219		prolog_code_error(Kind,Code) :- cur_line(L),
220		format(user_error,'Unexpected character for ~w on line ~w: ~w "~s"~n',[Kind,L,Code,[Code]]),
221		add_error(fastread,'Unexpected character code on line:',Kind:Code:line(L)).
222
223		:- dynamic cur_line/1.
224		cur_line(1).
225		inc_line_nr :- retract(cur_line(N)), N1 is N+1, assertz(cur_line(N1)).
226		reset_line_nr :- retract(cur_line(_)), assertz(cur_line(1)).
227
228		% ------------------------
229
230		:- assert_must_succeed(tools_fastread:read_term_from_string("atom('a',1,2).",atom(a,1,2))).
231		:- assert_must_succeed(tools_fastread:read_term_from_string("'\\''.",'\'')).
232		:- assert_must_succeed(tools_fastread:read_term_from_string("'\\374\\'.",'\374\')).
233
234		:- use_module(library(codesio),[open_codes_stream/2]).
235		read_term_from_string(Codes,Term) :-
236		open_codes_stream(Codes,Stream),
237		read_term_from_stream(Stream,Term),
238		close(Stream).
239
240
241		% read next term on stream, term has to be followed by '.'
242		read_term_from_stream(Stream,Term) :-
243		get_next_token(Stream,Token),!,
244		get_next_term(Token,Stream,Term),!,
245		(Term='end_of_file' -> true
246		; expect_token(Stream,dot)).
247
248		% get one term, without dot at end:
249		get_next_term(eof,_,'end_of_file').
250		get_next_term(number(C),_,Term) :- number_codes(Term,C).
251		get_next_term(atom(C),Stream,Term) :- atom_codes(Atom,C), complete_atom(Atom,Stream,Term).
252		get_next_term(open_bracket,Stream,List) :- get_term_list(Stream,close_bracket,List).
253		get_next_term(open_paren,Stream,'$pair'(List)) :- get_term_list(Stream,close_paren,List).
254		get_next_term(close_paren,_,_) :- prolog_token_error(dot),fail.
255		get_next_term(close_bracket,_,_) :- prolog_token_error(dot),fail.
256		get_next_term(comma,_,_) :- prolog_token_error(dot),fail.
257		get_next_term(dot,_,_) :- prolog_token_error(dot),fail.
258
259		% check if an atom is followed by a list of arguments (Term1,...)
260		complete_atom(Atom,Stream,Term) :- skip_ws(Stream,C), C==0'(, !,
261		get_code(Stream,_),
262		% TO DO: special case for binary operators like conjunct, transform into flat list (associativity holds) meaning we can get tail-recursion
263		get_term_list(Stream,close_paren,List),
264		Term =.. [Atom|List].
265		complete_atom(Atom,_,Atom). % atom without parentheses, i.e., on its own
266
267		% pre-construct term for certain functors to make above call tail-recursive
268		% however, as conjunctions are often stored the other way conjunct(pos,conjunct(...),Pred) this is not really beneficial
269		% we could have a special get_conjunct which creates a list
270		known_arity(empty_set,empty_set(Pos),[Pos]).
271		known_arity(boolean_true,boolean_true(Pos),[Pos]).
272		known_arity(boolean_false,boolean_false(Pos),[Pos]).
273		known_arity(identifier,identifier(Pos,ID),[Pos,ID]).
274		known_arity(integer,integer(Pos,I),[Pos,I]).
275		known_arity(string,string(Pos,I),[Pos,I]).
276		known_arity(equal,equal(Pos,A,B),[Pos,A,B]).
277		known_arity(not_equal,not_equal(Pos,A,B),[Pos,A,B]).
278		%known_arity(couple,couple(Pos,A,B),[Pos,A,B]). % couple(Pos,List)
279		known_arity(couple,couple(Pos,L),[Pos,L]).
280		known_arity(set_extension,set_extension(Pos,L),[Pos,L]).
281		known_arity(sequence_extension,sequence_extension(Pos,L),[Pos,L]).
282		known_arity(conjunct,conjunct(Pos,A,B),[Pos,A,B]).
283		known_arity(disjunct,disjunct(Pos,A,B),[Pos,A,B]).
284		known_arity(interval,interval(Pos,A,B),[Pos,A,B]).
285		known_arity(member,member(Pos,A,B),[Pos,A,B]).
286		known_arity(function,function(Pos,A,B),[Pos,A,B]).
287		known_arity(assign,assign(Pos,A,B),[Pos,A,B]).
288		known_arity(p3,p3(A,B,C),[A,B,C]).
289		known_arity(p4,p4(A,B,C,D),[A,B,C,D]).
290		known_arity(p5,p5(A,B,C,D,E),[A,B,C,D,E]).
291
292		% get list of terms separated by , and terminated by provided token
293		get_term_list(Stream,EndToken,List) :- get_next_token(Stream,Token), get_term_list_aux(Token,Stream,EndToken,List).
294
295		get_term_list_aux(EndToken,_,EndToken,List) :- !, List=[].
296		get_term_list_aux(Token,Stream,EndToken,[Term1|TList]) :- get_next_term(Token,Stream,Term1),
297		get_next_token(Stream,Token2),!,
298		get_term_list_aux2(Token2,Stream,EndToken,TList).
299
300		get_term_list_aux2(EndToken,_,EndToken,List) :- !, List=[].
301		get_term_list_aux2(comma,Stream,EndToken,[Term1|TList]) :-
302		get_next_token(Stream,Token1),
303		get_next_term(Token1,Stream,Term1),!,
304		get_next_token(Stream,Token2),
305		get_term_list_aux2(Token2,Stream,EndToken,TList).
306		get_term_list_aux2(Token2,_,EndToken,_) :-
307		format(user_error,'Unexpected token: ~w, expected ~w~n',[Token2,EndToken]),fail.
308
309		term_read(File) :- statistics(walltime,_),
310		open(File,read,Stream),read_terms(Stream),close(Stream),
311		statistics(walltime,[_,W1]), format('~n% Walltime ~w ms to tokenise and read ~w~n',[W1,File]),
312		print_memory_used_wo_gc.
313
314		read_terms(Stream) :-
315		read_term_from_stream(Stream,Term),
316		%read(Stream,Term), % Prolog's read
317		write_term(Term,[max_depth(5),numbervars(true)]),nl,
318		!,
319		(Term = 'end_of_file' -> true
320		; read_terms(Stream)).
321		read_terms(Stream) :-
322		get_code(Stream,C),
323		format('Could not process: ~w~n',[C]).
324
325		read_test(1) :-
326		term_read('/Users/leuschel/git_root/prob_examples/public_examples/B/Demo/Bakery0.prob').
327		read_test(3) :-
328		term_read('/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated4000.prob').
329		read_test(4) :-
330		term_read('/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.prob').
331
332		read_test(7) :-
333		term_read('/Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.prob').
334
335		read_test(8) :-
336		term_read('/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/SATLIB/sudoku.prob').
337
338
339
340		/*
341
342		with new read_term_from_stream:
343
344		| ?- read_test(4).
345		parser_version(2019-03-05 17:22:26.861)
346		classical_b(Generated10000,[/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.mch])
347		machine(abstract_machine(pos(1,1,1,1,50007,4),machine(pos(2,1,1,1,1,8)),machine_header(pos(3,1,1,9,1,23),Generated10000,[]),[constants(pos(...),[...|...]),properties(pos(...),conjunct(...)),variables(...)|...]))
348		end_of_file
349
350		% Walltime 774 ms to tokenise /Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.prob
351		108.476 MB
352
353		with fastreadwrite:
354		% Walltime 89 ms to fastread from ./fastrw_out.probz
355		109.004 MB
356
357
358		with Prolog read
359		| ?- rtest4.
360		parser_version(2019-03-05 17:22:26.861)
361		classical_b(Generated10000,[/Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.mch])
362		machine(abstract_machine(pos(1,1,1,1,50007,4),machine(pos(2,1,1,1,1,8)),machine_header(pos(3,1,1,9,1,23),Generated10000,[]),[constants(pos(...),[...|...]),properties(pos(...),conjunct(...)),variables(...)|...]))
363		end_of_file
364
365		% Walltime 4193 ms to tokenise /Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated10000.prob
366		286.429 MB
367
368
369		% Walltime 102 ms to get_codes /Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated4000.prob
370
371		% Walltime 233 ms to tokenise /Users/leuschel/git_root/prob_examples/public_examples/B/PerformanceTests/Generated/Generated4000.prob
372		116.813 MB ( 11.956 MB program)
373
374
375
376		| ?- rtest7.
377		parser_version(23928fd10a7a580ccd609bf614ed4552107bbd6a)
378		classical_b(rule,[/Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.mch,/Users/leuschel/git_root/prob_prolog/stdlib/CHOOSE.def,/Users/leuschel/git_root/prob_prolog/stdlib/LibraryBits.def,/Users/leuschel/git_root/prob_prolog/stdlib/LibraryCSV.def|...])
379		machine(abstract_machine(pos(4687857,17,1,1,79,4),machine(pos(4687858,17,1,1,1,8)),machine_header(pos(4687859,17,1,9,1,21),LibraryFiles,[]),[abstract_constants(pos(...),[...|...]),properties(pos(...),conjunct(...)),definitions(...)|...]))
380		machine(abstract_machine(pos(4688547,18,1,1,94,4),machine(pos(4688548,18,1,1,1,8)),machine_header(pos(4688549,18,1,9,1,20),LibraryMath,[]),[abstract_constants(pos(...),[...|...]),definitions(pos(...),[...|...]),properties(...)|...]))
381		machine(abstract_machine(pos(4689324,19,1,1,176,4),machine(pos(4689325,19,1,1,1,8)),machine_header(pos(4689326,19,1,9,1,23),LibraryStrings,[]),[abstract_constants(pos(...),[...|...]),properties(pos(...),conjunct(...)),definitions(...)|...]))
382		machine(abstract_machine(pos(1,1,1,1,1929826,4),machine(pos(2,1,1,1,1,8)),machine_header(pos(3,1,2,5,2,9),rule,[]),[definitions(pos(...),[...|...]),sees(pos(...),[...|...]),constants(...)|...]))
383		end_of_file
384
385		% Walltime 31964 ms to tokenise /Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.prob
386		659.087 MB
387		yes
388
389		% Walltime 6243 ms to fastread from ./fastrw_out.probz
390		659.025 MB
391
392
393		% Walltime 5676 ms to get_codes /Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.prob
394
395		% Walltime 15616 ms to tokenise /Users/leuschel/git_root/private_examples/ClearSy/2019_Nov/rule_perf4/rule_split.prob
396		116.840 MB ( 11.982 MB program)
397
398		*/
399