1 // This file contains most of the magic that extracts "context
2 // information" from the unstructered layout-oriented stuff in an
7 #include "math_amsarrayinset.h"
8 #include "math_arrayinset.h"
9 #include "math_charinset.h"
10 #include "math_deliminset.h"
11 #include "math_diffinset.h"
12 #include "math_exfuncinset.h"
13 #include "math_exintinset.h"
14 #include "math_fracinset.h"
15 #include "math_liminset.h"
16 #include "math_matrixinset.h"
17 #include "math_mathmlstream.h"
18 #include "math_numberinset.h"
19 #include "math_scriptinset.h"
20 #include "math_stringinset.h"
21 #include "math_symbolinset.h"
22 #include "math_unknowninset.h"
23 #include "math_parser.h"
26 #include "support/lyxlib.h"
27 #include "support/systemcall.h"
28 #include "support/filetools.h"
33 using std::istringstream;
38 ostream & operator<<(ostream & os, MathArray const & ar)
46 // define a function for tests
47 typedef bool TestItemFunc(MathAtom const &);
49 // define a function for replacing subexpressions
50 typedef MathAtom ReplaceArgumentFunc(const MathArray & ar);
54 // try to extract a super/subscript
55 // modify iterator position to point behind the thing
56 bool extractScript(MathArray & ar,
57 MathArray::iterator & pos, MathArray::iterator last)
59 // nothing to get here
63 // is this a scriptinset?
64 if (!(*pos)->asScriptInset())
67 // it is a scriptinset, use it.
74 // try to extract an "argument" to some function.
75 // returns position behind the argument
76 MathArray::iterator extractArgument(MathArray & ar,
77 MathArray::iterator pos, MathArray::iterator last, string const & = "")
79 // nothing to get here
83 // something deliminited _is_ an argument
84 if ((*pos)->asDelimInset()) {
89 // always take the first thing, no matter what it is
92 // go ahead if possible
97 // if the next item is a subscript, it most certainly belongs to the
99 extractScript(ar, pos, last);
103 // but it might be more than that.
104 // FIXME: not implemented
105 //for (MathArray::iterator it = pos + 1; it != last; ++it) {
106 // // always take the first thing, no matter
108 // ar.push_back(*it);
116 // returns sequence of char with same code starting at it up to end
117 // it might be less, though...
119 (MathArray::const_iterator it, MathArray::const_iterator end)
122 for (; it != end && (*it)->asCharInset(); ++it)
123 s += (*it)->getChar();
128 void extractStrings(MathArray & ar)
130 //lyxerr << "\nStrings from: " << ar << "\n";
131 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
132 if (!ar[i]->asCharInset())
134 string s = charSequence(ar.begin() + i, ar.end());
135 ar[i] = MathAtom(new MathStringInset(s));
136 ar.erase(i + 1, i + s.size());
138 //lyxerr << "\nStrings to: " << ar << "\n";
142 void extractMatrices(MathArray & ar)
144 //lyxerr << "\nMatrices from: " << ar << "\n";
145 // first pass for explicitly delimited stuff
146 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
147 if (!ar[i]->asDelimInset())
149 MathArray const & arr = ar[i]->asDelimInset()->cell(0);
152 if (!arr.front()->asGridInset())
154 ar[i] = MathAtom(new MathMatrixInset(*(arr.front()->asGridInset())));
157 // second pass for AMS "pmatrix" etc
158 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
159 if (ar[i]->asAMSArrayInset())
161 ar[i] = MathAtom(new MathMatrixInset(*(ar[i]->asGridInset())));
163 //lyxerr << "\nMatrices to: " << ar << "\n";
167 // convert this inset somehow to a string
168 bool extractString(MathAtom const & at, string & str)
171 str = string(1, at->getChar());
174 if (at->asStringInset()) {
175 str = at->asStringInset()->str();
182 // convert this inset somehow to a number
183 bool extractNumber(MathArray const & ar, int & i)
185 istringstream is(charSequence(ar.begin(), ar.end()).c_str());
191 bool extractNumber(MathArray const & ar, double & d)
193 istringstream is(charSequence(ar.begin(), ar.end()).c_str());
199 bool testString(MathAtom const & at, const string & str)
202 return extractString(at, s) && str == s;
206 // search end of nested sequence
207 MathArray::iterator endNestSearch(
208 MathArray::iterator it,
209 MathArray::iterator last,
210 TestItemFunc testOpen,
211 TestItemFunc testClose
214 for (int level = 0; it != last; ++it) {
226 // replace nested sequences by a real Insets
229 TestItemFunc testOpen,
230 TestItemFunc testClose,
231 ReplaceArgumentFunc replaceArg
234 // use indices rather than iterators for the loop because we are going
235 // to modify the array.
236 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
237 // check whether this is the begin of the sequence
238 if (!testOpen(ar[i]))
241 // search end of sequence
242 MathArray::iterator it = ar.begin() + i;
243 MathArray::iterator jt = endNestSearch(it, ar.end(), testOpen, testClose);
247 // replace the original stuff by the new inset
248 ar.erase(it + 1, jt + 1);
250 // create a proper inset as replacement
251 ar[i] = replaceArg(MathArray(it + 1, jt));
258 // split scripts into seperate super- and subscript insets. sub goes in
262 void splitScripts(MathArray & ar)
264 //lyxerr << "\nScripts from: " << ar << "\n";
265 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
266 // is this script inset?
267 if (!ar[i]->asScriptInset())
270 // no problem if we don't have both...
271 if (!ar[i]->asScriptInset()->hasUp())
273 if (!ar[i]->asScriptInset()->hasDown())
276 // create extra script inset and move superscript over
277 MathScriptInset * p = ar[i].nucleus()->asScriptInset();
278 MathScriptInset * q = new MathScriptInset(true);
279 std::swap(q->up(), p->up());
280 p->removeScript(true);
282 // insert new inset behind
284 ar.insert(i, MathAtom(q));
286 //lyxerr << "\nScripts to: " << ar << "\n";
294 void extractExps(MathArray & ar)
296 //lyxerr << "\nExps from: " << ar << "\n";
297 for (MathArray::size_type i = 0; i + 1 < ar.size(); ++i) {
299 if (ar[i]->getChar() != 'e')
302 // we need an exponent but no subscript
303 MathScriptInset const * sup = ar[i + 1]->asScriptInset();
304 if (!sup || sup->hasDown())
307 // create a proper exp-inset as replacement
308 ar[i] = MathAtom(new MathExFuncInset("exp", sup->cell(1)));
311 //lyxerr << "\nExps to: " << ar << "\n";
316 // extract det(...) from |matrix|
318 void extractDets(MathArray & ar)
320 //lyxerr << "\ndet from: " << ar << "\n";
321 for (MathArray::iterator it = ar.begin(); it != ar.end(); ++it) {
322 MathDelimInset const * del = (*it)->asDelimInset();
327 *it = MathAtom(new MathExFuncInset("det", del->cell(0)));
329 //lyxerr << "\ndet to: " << ar << "\n";
337 bool isDigitOrSimilar(char c)
339 return ('0' <= c && c <= '9') || c == '.';
343 // returns sequence of digits
345 (MathArray::const_iterator it, MathArray::const_iterator end)
348 for (; it != end && (*it)->asCharInset(); ++it) {
349 if (!isDigitOrSimilar((*it)->getChar()))
351 s += (*it)->getChar();
357 void extractNumbers(MathArray & ar)
359 //lyxerr << "\nNumbers from: " << ar << "\n";
360 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
361 if (!ar[i]->asCharInset())
363 if (!isDigitOrSimilar(ar[i]->asCharInset()->getChar()))
366 string s = digitSequence(ar.begin() + i, ar.end());
368 ar[i] = MathAtom(new MathNumberInset(s));
369 ar.erase(i + 1, i + s.size());
371 //lyxerr << "\nNumbers to: " << ar << "\n";
377 // search deliminiters
380 bool testOpenParan(MathAtom const & at)
382 return testString(at, "(");
386 bool testCloseParan(MathAtom const & at)
388 return testString(at, ")");
392 MathAtom replaceDelims(const MathArray & ar)
394 return MathAtom(new MathDelimInset("(", ")", ar));
398 // replace '('...')' sequences by a real MathDelimInset
399 void extractDelims(MathArray & ar)
401 //lyxerr << "\nDelims from: " << ar << "\n";
402 replaceNested(ar, testOpenParan, testCloseParan, replaceDelims);
403 //lyxerr << "\nDelims to: " << ar << "\n";
409 // search well-known functions
413 // replace 'f' '(...)' and 'f' '^n' '(...)' sequences by a real MathExFuncInset
414 // assume 'extractDelims' ran before
415 void extractFunctions(MathArray & ar)
417 // we need at least two items...
421 //lyxerr << "\nFunctions from: " << ar << "\n";
422 for (MathArray::size_type i = 0; i + 1 < ar.size(); ++i) {
423 MathArray::iterator it = ar.begin() + i;
424 MathArray::iterator jt = it + 1;
428 if ((*it)->asUnknownInset()) {
429 // it certainly is if it is well known...
430 name = (*it)->name();
432 // is this a user defined function?
433 // it it probably not, if it doesn't have a name.
434 if (!extractString(*it, name))
436 // it is not if it has no argument
439 // guess so, if this is followed by
440 // a DelimInset with a single item in the cell
441 MathDelimInset const * del = (*jt)->asDelimInset();
442 if (!del || del->cell(0).size() != 1)
444 // fall trough into main branch
447 // do we have an exponent like in
448 // 'sin' '^2' 'x' -> 'sin(x)' '^2'
450 extractScript(exp, jt, ar.end());
452 // create a proper inset as replacement
453 MathExFuncInset * p = new MathExFuncInset(name);
455 // jt points to the "argument". Get hold of this.
456 MathArray::iterator st = extractArgument(p->cell(0), jt, ar.end());
458 // replace the function name by a real function inset
461 // remove the source of the argument from the array
462 ar.erase(it + 1, st);
464 // re-insert exponent
465 ar.insert(i + 1, exp);
466 //lyxerr << "\nFunctions to: " << ar << "\n";
475 bool testSymbol(MathAtom const & at, string const & name)
477 return at->asSymbolInset() && at->asSymbolInset()->name() == name;
481 bool testIntSymbol(MathAtom const & at)
483 return testSymbol(at, "int");
487 bool testIntegral(MathAtom const & at)
491 ( at->asScriptInset()
492 && at->asScriptInset()->nuc().size()
493 && testIntSymbol(at->asScriptInset()->nuc().back()) );
498 bool testIntDiff(MathAtom const & at)
500 return testString(at, "d");
504 // replace '\int' ['_^'] x 'd''x'(...)' sequences by a real MathExIntInset
505 // assume 'extractDelims' ran before
506 void extractIntegrals(MathArray & ar)
508 // we need at least three items...
512 //lyxerr << "\nIntegrals from: " << ar << "\n";
513 for (MathArray::size_type i = 0; i + 1 < ar.size(); ++i) {
514 MathArray::iterator it = ar.begin() + i;
517 MathArray::iterator jt =
518 endNestSearch(it, ar.end(), testIntegral, testIntDiff);
520 // something sensible found?
524 // is this a integral name?
525 if (!testIntegral(*it))
528 // core ist part from behind the scripts to the 'd'
529 MathExIntInset * p = new MathExIntInset("int");
531 // handle scripts if available
532 if (!testIntSymbol(*it)) {
533 p->cell(2) = (*it)->asScriptInset()->down();
534 p->cell(3) = (*it)->asScriptInset()->up();
536 p->cell(0) = MathArray(it + 1, jt);
538 // use the "thing" behind the 'd' as differential
539 MathArray::iterator tt = extractArgument(p->cell(1), jt + 1, ar.end());
542 ar.erase(it + 1, tt);
545 //lyxerr << "\nIntegrals to: " << ar << "\n";
554 bool testEqualSign(MathAtom const & at)
556 return testString(at, "=");
560 bool testSumSymbol(MathAtom const & p)
562 return testSymbol(p, "sum");
566 bool testSum(MathAtom const & at)
570 ( at->asScriptInset()
571 && at->asScriptInset()->nuc().size()
572 && testSumSymbol(at->asScriptInset()->nuc().back()) );
576 // replace '\sum' ['_^'] f(x) sequences by a real MathExIntInset
577 // assume 'extractDelims' ran before
578 void extractSums(MathArray & ar)
580 // we need at least two items...
584 //lyxerr << "\nSums from: " << ar << "\n";
585 for (MathArray::size_type i = 0; i + 1 < ar.size(); ++i) {
586 MathArray::iterator it = ar.begin() + i;
588 // is this a sum name?
592 // create a proper inset as replacement
593 MathExIntInset * p = new MathExIntInset("sum");
595 // collect lower bound and summation index
596 MathScriptInset const * sub = ar[i]->asScriptInset();
597 if (sub && sub->hasDown()) {
598 // try to figure out the summation index from the subscript
599 MathArray const & ar = sub->down();
600 MathArray::const_iterator xt =
601 find_if(ar.begin(), ar.end(), &testEqualSign);
602 if (xt != ar.end()) {
603 // we found a '=', use everything in front of that as index,
604 // and everything behind as lower index
605 p->cell(1) = MathArray(ar.begin(), xt);
606 p->cell(2) = MathArray(xt + 1, ar.end());
608 // use everything as summation index, don't use scripts.
613 // collect upper bound
614 if (sub && sub->hasUp())
615 p->cell(3) = sub->up();
617 // use something behind the script as core
618 MathArray::iterator tt = extractArgument(p->cell(0), it + 1, ar.end());
621 ar.erase(it + 1, tt);
624 //lyxerr << "\nSums to: " << ar << "\n";
629 // search differential stuff
632 // tests for 'd' or '\partial'
633 bool testDiffItem(MathAtom const & at)
635 return testString(at, "d");
639 bool testDiffArray(MathArray const & ar)
641 return ar.size() && testDiffItem(ar.front());
645 bool testDiffFrac(MathAtom const & at)
649 && testDiffArray(at->asFracInset()->cell(0))
650 && testDiffArray(at->asFracInset()->cell(1));
654 void extractDiff(MathArray & ar)
656 //lyxerr << "\nDiffs from: " << ar << "\n";
657 for (MathArray::size_type i = 0; i < ar.size(); ++i) {
658 MathArray::iterator it = ar.begin() + i;
660 // is this a "differential fraction"?
661 if (!testDiffFrac(*it))
664 MathFracInset const * f = (*it)->asFracInset();
666 lyxerr << "should not happen\n";
670 // create a proper diff inset
671 MathDiffInset * diff = new MathDiffInset;
673 // collect function, let jt point behind last used item
674 MathArray::iterator jt = it + 1;
676 MathArray const & numer = f->cell(0);
677 if (numer.size() > 1 && numer[1]->asScriptInset()) {
678 // this is something like d^n f(x) / d... or d^n / d...
681 if (numer.size() > 2)
682 diff->cell(0) = MathArray(numer.begin() + 2, numer.end());
684 jt = extractArgument(diff->cell(0), jt, ar.end());
686 // simply d f(x) / d... or d/d...
687 if (numer.size() > 1)
688 diff->cell(0) = MathArray(numer.begin() + 1, numer.end());
690 jt = extractArgument(diff->cell(0), jt, ar.end());
693 // collect denominator parts
694 MathArray const & denom = f->cell(1);
695 for (MathArray::const_iterator dt = denom.begin(); dt != denom.end();) {
697 MathArray::const_iterator et
698 = find_if(dt + 1, denom.end(), &testDiffItem);
701 MathArray::const_iterator st = et - 1;
702 MathScriptInset const * script = (*st)->asScriptInset();
703 if (script && script->hasUp()) {
704 // things like d.../dx^n
706 if (extractNumber(script->up(), mult)) {
707 //lyxerr << "mult: " << mult << endl;
708 for (int i = 0; i < mult; ++i)
709 diff->addDer(MathArray(dt + 1, st));
713 diff->addDer(MathArray(dt + 1, et));
719 ar.erase(it + 1, jt);
720 *it = MathAtom(diff);
722 //lyxerr << "\nDiffs to: " << ar << "\n";
731 bool testRightArrow(MathAtom const & at)
733 return testSymbol(at, "to") || testSymbol(at, "rightarrow");
738 // replace '\lim_{x->x0} f(x)' sequences by a real MathLimInset
739 // assume 'extractDelims' ran before
740 void extractLims(MathArray & ar)
742 // we need at least three items...
746 //lyxerr << "\nLimits from: " << ar << "\n";
747 for (MathArray::size_type i = 0; i + 2 < ar.size(); ++i) {
748 MathArray::iterator it = ar.begin() + i;
750 // is this a limit function?
751 if (!testSymbol(*it, "lim"))
754 // the next one must be a subscript (without superscript)
755 MathScriptInset const * sub = (*(it + 1))->asScriptInset();
756 if (!sub || !sub->hasDown() || sub->hasUp())
759 // and it must contain a -> symbol
760 MathArray const & s = sub->down();
761 MathArray::const_iterator st = find_if(s.begin(), s.end(), &testRightArrow);
765 // the -> splits the subscript int x and x0
766 MathArray x = MathArray(s.begin(), st);
767 MathArray x0 = MathArray(st + 1, s.end());
769 // use something behind the script as core
771 MathArray::iterator tt = extractArgument(f, it + 2, ar.end());
774 ar.erase(it + 1, tt);
776 // create a proper inset as replacement
777 *it = MathAtom(new MathLimInset(f, x, x0));
779 //lyxerr << "\nLimits to: " << ar << "\n";
787 void extractStructure(MathArray & ar)
789 extractIntegrals(ar);
795 extractFunctions(ar);
804 void write(MathArray const & dat, WriteStream & wi)
808 wi.firstitem() = true;
809 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it) {
811 wi.firstitem() = false;
816 void normalize(MathArray const & ar, NormalStream & os)
818 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it)
819 (*it)->normalize(os);
823 void octavize(MathArray const & dat, OctaveStream & os)
826 extractStructure(ar);
827 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it)
832 void maplize(MathArray const & dat, MapleStream & os)
835 extractStructure(ar);
836 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it)
841 void mathematicize(MathArray const & dat, MathematicaStream & os)
844 extractStructure(ar);
845 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it)
846 (*it)->mathematicize(os);
850 void mathmlize(MathArray const & dat, MathMLStream & os)
853 extractStructure(ar);
856 else if (ar.size() == 1)
860 for (MathArray::const_iterator it = ar.begin(); it != ar.end(); ++it)
861 (*it)->mathmlize(os);
871 string captureOutput(string const & cmd, string const & data)
873 string outfile = lyx::tempName(string(), "mathextern");
874 string full = "echo '" + data + "' | (" + cmd + ") > " + outfile;
875 lyxerr << "calling: " << full << endl;
877 dummy.startscript(Systemcall::Wait, full);
878 string out = GetFileContents(outfile);
879 lyx::unlink(outfile);
880 lyxerr << "result: '" << out << "'" << endl;
885 MathArray pipeThroughMaple(string const & extra, MathArray const & ar)
887 string header = "readlib(latex):\n";
889 // remove the \\it for variable names
890 //"#`latex/csname_font` := `\\it `:"
892 "`latex/csname_font` := ``:\n";
894 // export matrices in (...) instead of [...]
896 "`latex/latex/matrix` := "
897 "subs(`[`=`(`, `]`=`)`,"
898 "eval(`latex/latex/matrix`)):\n";
900 // replace \\cdots with proper '*'
902 "`latex/latex/*` := "
903 "subs(`\\,`=`\\cdot `,"
904 "eval(`latex/latex/*`)):\n";
906 // remove spurious \\noalign{\\medskip} in matrix output
908 "`latex/latex/matrix`:= "
909 "subs(`\\\\\\\\\\\\noalign{\\\\medskip}` = `\\\\\\\\`,"
910 "eval(`latex/latex/matrix`)):\n";
912 //"#`latex/latex/symbol` "
913 // " := subs((\\'_\\' = \\'`\\_`\\',eval(`latex/latex/symbol`)): ";
915 string trailer = "quit;";
919 string expr = os.str().c_str();
920 lyxerr << "ar: '" << ar << "'\n";
921 lyxerr << "ms: '" << os.str() << "'\n";
923 for (int i = 0; i < 100; ++i) { // at most 100 attempts
924 // try to fix missing '*' the hard way by using mint
926 // ... > echo "1A;" | mint -i 1 -S -s -q
929 // Probably missing an operator such as * p
931 lyxerr << "checking expr: '" << expr << "'\n";
932 string out = captureOutput("mint -i 1 -S -s -q -q", expr + ";");
934 break; // expression syntax is ok
935 istringstream is(out.c_str());
938 if (line.find("on line") != 0)
939 break; // error message not identified
941 string::size_type pos = line.find('^');
942 if (pos == string::npos || pos < 15)
943 break; // caret position not found
944 pos -= 15; // skip the "on line ..." part
945 if (expr[pos] == '*' || (pos > 0 && expr[pos - 1] == '*'))
946 break; // two '*' in a row are definitely bad
947 expr.insert(pos, "*");
950 string full = "latex(" + extra + '(' + expr + "));";
951 string out = captureOutput("maple -q", header + full + trailer);
957 mathed_parse_cell(res, out);
962 MathArray pipeThroughOctave(string const &, MathArray const & ar)
967 string expr = os.str().c_str();
970 lyxerr << "pipe: ar: '" << ar << "'\n";
971 lyxerr << "pipe: expr: '" << expr << "'\n";
973 for (int i = 0; i < 100; ++i) { // at most 100 attempts
975 // try to fix missing '*' the hard way
977 // >>> ([[1 2 3 ];[2 3 1 ];[3 1 2 ]])([[1 2 3 ];[2 3 1 ];[3 1 2 ]])
980 lyxerr << "checking expr: '" << expr << "'\n";
981 out = captureOutput("octave -q 2>&1", expr);
982 lyxerr << "checking out: '" << out << "'\n";
984 // leave loop if expression syntax is probably ok
985 if (out.find("parse error:") == string::npos)
988 // search line with single caret
989 istringstream is(out.c_str());
993 lyxerr << "skipping line: '" << line << "'\n";
994 if (line.find(">>> ") != string::npos)
998 // found line with error, next line is the one with caret
1000 string::size_type pos = line.find('^');
1001 lyxerr << "caret line: '" << line << "'\n";
1002 lyxerr << "found caret at pos: '" << pos << "'\n";
1003 if (pos == string::npos || pos < 4)
1004 break; // caret position not found
1005 pos -= 4; // skip the ">>> " part
1006 if (expr[pos] == '*')
1007 break; // two '*' in a row are definitely bad
1008 expr.insert(pos, "*");
1015 out = out.substr(6);
1017 // parse output as matrix or single number
1018 MathAtom at(new MathArrayInset("array", out));
1019 MathArrayInset const * mat = at->asArrayInset();
1021 if (mat->ncols() == 1 && mat->nrows() == 1)
1022 res.append(mat->cell(0));
1024 res.push_back(MathAtom(new MathDelimInset("(", ")")));
1025 res.back().nucleus()->cell(0).push_back(at);
1033 MathArray pipeThroughExtern(string const & lang, string const & extra,
1034 MathArray const & ar)
1036 if (lang == "octave")
1037 return pipeThroughOctave(extra, ar);
1039 if (lang == "maple")
1040 return pipeThroughMaple(extra, ar);
1042 // create normalized expression
1044 NormalStream ns(os);
1045 os << "[" << extra << ' ';
1048 string data = os.str().c_str();
1050 // search external script
1051 string file = LibFileSearch("mathed", "extern_" + lang);
1053 lyxerr << "converter to '" << lang << "' not found\n";
1057 // run external sript
1058 string out = captureOutput(file, data);
1060 mathed_parse_cell(res, out);