sources-request@panda.UUCP (08/23/85)
Mod.sources: Volume 3, Issue 2
Submitted by: Peter Bain <ihnp4!watmath!wateng!pdbain>
Match is functionally identical to bm, but uses the Vax MATCHC instruction
to do the pattern matcher. On 780s, 785s, probably Microvax 2s, 730s
and 750s (but NOT microvax 1s) it runs considerably faster. Here is
a run on our 780 under 4.2 BSD:
/bin/time grep foobar /usr/dict/words
16.7 real 4.6 user 0.6 sys
/bin/time bm foobar /usr/dict/words
3.0 real 0.9 user 0.5 sys
/bin/time match foobar /usr/dict/words
3.5 real 0.3 user 0.5 sys
Note that it runs ONLY on Vaxen, and possibly only under 4.2.
------------------ tear here ----------------------------------
: This is a shar archive. Extract with sh, not csh.
: The rest of this file will extract:
: Execute.c Extern.h GetPatFile.c Global.c MakeDesc.c match.h match.c MatchFound.c MkDescVec.c MoveResidue.c PrintLine.c PutUsage.c Search.s Makefile README match.p
echo Extracting Execute.c
sed 's/^X//' > Execute.c << 'e-o-f'
X#include <stdio.h>
X#include "match.h"
X#include "Extern.h"
XExecute(DescVec, NPats, TextFile, Buffer)
Xregister struct PattDesc *DescVec[];
X/* pointers to status vectors for the different
X * patterns, including skip tables, position in buffer, etc. */
Xint NPats; /* number of patterns */
Xregister char Buffer[]; /* holds text from file */
Xint TextFile; /* file to search */
X{
X int NRead, /* number of chars read from file */
X NWanted, /* number of chars wanted */
X NAvail, /* number of chars actually read */
X BuffSize, /* number of chars in buffer */
X BuffPos, /* offset of first char in Buffer in TextFile */
X BuffEx, /* flag to indicate that buffer has been searched */
X ResSize,
X /* number of characters in the last, incomplete line */
X Found, /* flag indicates whether pattern found
X * completely and all matches printed */
X Valid; /* was the match "valid", i.e. if -x used,
X * did the whole line match? */
X register char *BuffEnd;
X /* pointer to last char of last complete line */
X
X /* misc working variables */
X register int i;
X
X /* initialize */
X ResSize = 0;
X Found = 0;
X BuffPos = 0;
X for (i=0; i < NPats; i++) {
X DescVec[i] -> Success = 0;
X DescVec[i] -> Start = Buffer;
X } /* for */
X /* now do the searching */
X do {
X /* first, read a bufferfull and set up the variables */
X NWanted = MAXBUFF - ResSize; NRead = 0;
X do {
X NAvail =
X read(TextFile,Buffer + ResSize + NRead, NWanted);
X if (NAvail == -1) {
X fprintf(stderr,
X "bm: error reading from input file\n");
X exit(2);
X } /* if */
X NRead += NAvail; NWanted -= NAvail;
X } while (NAvail && NWanted);
X BuffEx = 0;
X BuffSize = ResSize + NRead;
X BuffEnd = Buffer + BuffSize - 1;
X /* locate the end of the last complete line */
X while (*BuffEnd != '\n' && BuffEnd >= Buffer)
X --BuffEnd;
X if (BuffEnd < Buffer)
X BuffEnd = Buffer + BuffSize - 1;
X while (!BuffEx) { /* work through one buffer full */
X BuffEx = 1; /* set it provisionally, then clear
X * it if we find the buffer non-empty */
X for (i=0; i< NPats; i++) {
X if (!DescVec[i]->Success)
X /* if the pattern has not been found */
X DescVec[i]-> Success =
X Search(DescVec[i]->Pattern,
X DescVec[i]->PatLen, DescVec[i]->Start,
X BuffEnd - DescVec[i]->Start + 1,
X DescVec[i]);
X if (DescVec[i]->Success){
X /* if a match occurred */
X BuffEx = 0;
X Valid = MatchFound(DescVec[i],BuffPos,
X Buffer, BuffEnd);
X Found |= Valid;
X if ((sFlag || lFlag) && Found)
X return(0);
X } /* if */
X } /* for */
X } /* while */
X if(NRead) {
X ResSize = MoveResidue(DescVec,NPats,Buffer,
X Buffer + BuffSize -1);
X BuffPos += BuffSize - ResSize;
X } /* if */
X } while (NRead);
X return(!Found);
X} /* Execute */
e-o-f
echo Extracting Extern.h
sed 's/^X//' > Extern.h << 'e-o-f'
X/* global flags for bm */
Xextern int cFlag, /* true if we want only a count of matching lines */
X eFlag, /* indicates that next argument is the pattern */
X fFlag, /* true if the patterns arew to come from a file */
X lFlag, /* true if we want a list of files containing the pattern */
X nFlag, /* true if we want the character offset of the pattern */
X sFlag, /* true if we want silent mode */
X xFlag, /* true if we want only lines which match entirely */
X hFlag, /* true if we want no filenames in output */
X
X MatchCount; /* count of number of times a search string was found
X * in the text */
Xextern char *FileName;
e-o-f
echo Extracting GetPatFile.c
sed 's/^X//' > GetPatFile.c << 'e-o-f'
X#include <stdio.h>
X#include <sys/types.h>
X#include <sys/stat.h>
X#include "match.h"
Xint
XGetPatFile(PatFile, DescVec)
Xchar *PatFile;
Xstruct PattDesc *DescVec[];
X/* read patterns from a file and set up a pattern descriptor vector */
X{
X extern char *malloc();
X FILE *PFile;
X struct stat StatBuff;
X int PatSize; /* the number of chars in all the patterns */
X char *PatBuff; /* hold the patterns */
X if (!(PFile = fopen(PatFile,"r"))) {
X fprintf(stderr,"match: can't open pattern file %s\n",PatFile);
X exit(2);
X } /* if */
X /* find out how big the patterns are */
X if (fstat(fileno(PFile),&StatBuff) == -1) {
X fprintf(stderr,"match: can't fstat %s\n",PatFile);
X exit(2);
X } /* if */
X PatSize = StatBuff.st_size;
X if (!PatSize) {
X fprintf(stderr,"match: pattern file is empty\n");
X exit(2);
X } /* if */
X if (!(PatBuff = malloc(PatSize))) {
X fprintf(stderr,"match: insufficient memory to store patterns\n");
X exit(2);
X } /* if */
X fread(PatBuff,1,PatSize,PFile); /* get the patterns */
X /* make sure the patterns are null-terminated. We can't have
X * nulls in the patterns */
X if (PatBuff[PatSize-1] == '\n')
X PatBuff[PatSize-1] = '\0';
X else
X PatBuff[PatSize] = '\0';
X return(MkDescVec(DescVec,PatBuff));
X} /* GetPatFile */
e-o-f
echo Extracting Global.c
sed 's/^X//' > Global.c << 'e-o-f'
X/* global flags for match */
Xint cFlag=0, /* true if we want only a count of matching lines */
X eFlag=0, /* indicates that next argument is the pattern */
X fFlag=0, /* true if the patterns are to come from a file */
X lFlag=0, /* true if we want a list of files containing the pattern */
X nFlag=0, /* true if we want the character offset of the pattern */
X sFlag=0, /* true if we want silent mode */
X xFlag=0, /* true if we want only lines which match entirely */
X hFlag=0, /* true if we want no filenames in output */
X
X MatchCount=0; /* count of number of times a search string was found
X * in the text */
Xchar *FileName = 0;
e-o-f
echo Extracting MakeDesc.c
sed 's/^X//' > MakeDesc.c << 'e-o-f'
X#include <stdio.h>
X#include "match.h"
X#include "Extern.h"
Xextern char * malloc();
X/* makes a pattern descriptor */
Xstruct PattDesc *MakeDesc(Pattern)
Xchar *Pattern;
X{
X struct PattDesc *Desc;
X Desc = (struct PattDesc *) malloc(sizeof(struct PattDesc));
X Desc->Pattern=Pattern;
X Desc->PatLen = strlen(Desc->Pattern);
X return(Desc);
X} /* main */
e-o-f
echo Extracting match.h
sed 's/^X//' > match.h << 'e-o-f'
X#define FIRSTCHAR ' '
X#define MAXCHAR 0377
X#define MAXBUFF 8192
X#define MAXSIZE 100
X#define MAXPATS 100 /* max number of patterns */
X#define min(x,y) (x) < (y) ? (x) : (y)
X#define max(x,y) (x) > (y) ? (x) : (y)
Xstruct PattDesc {
X char *Pattern;
X int PatLen; /* pattern length */
X char *Start; /* starting position of search (at beginning of pattern) */
X int Success; /* true when pattern found */
X};
e-o-f
echo Extracting match.c
sed 's/^X//' > match.c << 'e-o-f'
X#include <stdio.h>
X#include <sys/file.h>
X#include <strings.h>
X#include "match.h"
X#include "Extern.h"
Xmain(argc,argv)
Xint argc;
Xchar *argv[];
X{
X /* test driver for grep based on Boyer-Moore algorithm */
X char BigBuff[MAXBUFF + 2];
X /*
X * We leave one extra character at the beginning and end of the buffer,
X * but don't tell Execute about it. This is so when someone is
X * scanning the buffer and scans past the end (or beginning)
X * we are still technically in the buffer (picky, but there ARE
X * machines which would complain)
X */
X int ret = 1, /* return code from Execute */
X NotFound = 0, /* non-zero if file not readable */
X NFiles,
X NPats; /* number of patterns to search for */
X char i,
X *FlagPtr,
X **OptPtr; /* used to scan command line */
X int TextFile /* file to search */;
X struct PattDesc *DescVec[MAXPATS];
X
X --argc;
X OptPtr = argv + 1;
X while ( argc && **OptPtr == '-') {
X FlagPtr = *OptPtr + 1;
X while (*FlagPtr) {
X switch (*FlagPtr) {
X case 'c': cFlag = 1; break;
X case 'e': eFlag = 1;
X /* get the patterns from next arg */
X NPats = MkDescVec(DescVec,*++OptPtr);
X --argc;
X break;
X case 'f': fFlag = 1;
X /* read the patterns from a file */
X NPats = GetPatFile(*++OptPtr,DescVec);
X --argc;
X break;
X case 'l': lFlag = 1; break;
X case 'n': nFlag = 1; break;
X case 's': sFlag = 1; break;
X case 'x': xFlag = 1; break;
X case 'h': hFlag = 1; break;
X default:
X fprintf(stderr,
X "match: invalid option: -%c \n",
X *FlagPtr);
X PutUsage();
X exit(2);
X } /* switch */
X ++FlagPtr;
X } /* while */
X ++OptPtr; --argc;
X } /* while */
X /* OptPtr now points to patterns */
X if (!fFlag && !eFlag) {
X if (!argc) {
X fprintf(stderr,"match: no pattern specified\n");
X PutUsage();
X exit(2);
X } else
X NPats = MkDescVec(DescVec,*OptPtr);
X ++OptPtr; --argc;
X }
X /* OptPtr now points to first file */
X NFiles = argc;
X if (!NFiles)
X ret &= Execute(DescVec,NPats,0,BigBuff+1);
X else while (argc--) {
X if ((NFiles > 1) || lFlag) FileName = *OptPtr;
X if ((TextFile = open(*OptPtr,O_RDONLY,0)) < 0) {
X fprintf(stderr,"match: can't open %s\n",*OptPtr);
X NotFound++;
X } else {
X ret &= Execute(DescVec,NPats,TextFile,BigBuff+1);
X if (sFlag && !ret)
X exit(0);
X close(TextFile);
X } /* if */
X ++OptPtr;
X } /* while */
X if (cFlag) printf("%d\n",MatchCount);
X exit(NotFound ? 2 : ret);
X} /* main */
e-o-f
echo Extracting MatchFound.c
sed 's/^X//' > MatchFound.c << 'e-o-f'
X#include <stdio.h>
X#include "match.h"
X#include "Extern.h"
XMatchFound(Desc, BuffPos, Buffer, BuffEnd)
Xstruct PattDesc *Desc; /* state info about search for one string */
Xint BuffPos; /* offset of first char of buffer into the file being searched */
Xchar *Buffer, /* pointer to the first character in the buffer */
X *BuffEnd; /* pointer to the last character in the buffer */
X{
X register char *MLineBegin, *MLineEnd;
X
X Desc->Success = 0;
X /* Start points to first character after a successful match */
X MLineBegin = MLineEnd = Desc->Start - 1;
X while(MLineBegin >=Buffer && *MLineBegin != '\n') --MLineBegin;
X ++MLineBegin;
X while( MLineEnd <= BuffEnd && *MLineEnd != '\n') ++MLineEnd;
X if (MLineEnd > BuffEnd) --MLineEnd;
X /* fixed 25jun85 pdbain. suppress multiple matches of the same
X * pattern on one line */
X Desc->Start = MLineEnd + 1;
X /* check if exact match */
X if (xFlag && !( Desc->PatLen == (*MLineEnd != '\n' ? ((MLineEnd -
X MLineBegin) + 1) : (MLineEnd - MLineBegin))))
X return(0); /* failure */
X if (sFlag) return(1);
X if (cFlag) {
X ++MatchCount;
X return(1);
X } /* if */
X PrintLine(BuffPos+(Desc->Start-Buffer),MLineBegin,MLineEnd);
X return(1);
X} /* MatchFound */
e-o-f
echo Extracting MkDescVec.c
sed 's/^X//' > MkDescVec.c << 'e-o-f'
X#include "match.h"
X#include <strings.h>
X/* scan a newline-separated string of patterns and set up the
X* vector of descriptors, one pattern descriptor per pattern.
X* Return the number of patterns */
Xint
XMkDescVec(DescVec, Pats)
Xstruct PattDesc *DescVec[];
Xchar *Pats;
X{
X int NPats = 0;
X char *EndPat;
X extern struct PattDesc *MakeDesc();
X while (*Pats && (EndPat = index(Pats,'\n')) && NPats < MAXPATS) {
X *EndPat = '\0';
X DescVec[NPats] = MakeDesc(Pats);
X Pats = EndPat + 1;
X ++NPats;
X } /* while */
X if (*Pats && NPats < MAXPATS) {
X DescVec[NPats] = MakeDesc(Pats);
X ++NPats;
X } /* if */
X return(NPats);
X} /* MkDescVec */
e-o-f
echo Extracting MoveResidue.c
sed 's/^X//' > MoveResidue.c << 'e-o-f'
X#include "match.h"
X/* Moves the text which has not been completely searched at the end of
X* the buffer to the beginning of the buffer
X* and returns number of bytes moved */
Xint MoveResidue(DescVec, NPats,Buffer, BuffEnd)
Xregister struct PattDesc **DescVec;
Xint NPats;
Xchar *Buffer, *BuffEnd;
X{
X char *FirstStart;
X register char *Residue;
X register int i;
X int ResSize;
X
X FirstStart = BuffEnd;
X /* find the earliest point which has not been
X * completely searched */
X for (i=0; i < NPats; i++) {
X FirstStart =
X min(FirstStart, DescVec[i]-> Start );
X } /* for */
X /* now scan to the beginning of the line containing the
X * unsearched text */
X for (Residue = FirstStart; *Residue != '\n' &&
X Residue >= Buffer; Residue--);
X if (Residue < Buffer)
X Residue = FirstStart;
X else ++Residue;
X /* now move the residue to the beginning of
X * the file */
X ResSize = BuffEnd - Residue + 1;
X bcopy(Residue, Buffer, ResSize);
X /* now fix the status vectors */
X for (i=0; i < NPats; i++) {
X DescVec[i]->Start -= (Residue - Buffer);
X } /* for */
X return(ResSize);
X} /* MoveResidue */
e-o-f
echo Extracting PrintLine.c
sed 's/^X//' > PrintLine.c << 'e-o-f'
X#include <stdio.h>
X#include <strings.h>
X#include "Extern.h"
XPrintLine(OffSet,LineStart,LineEnd)
Xint OffSet; /* offset of LineStart from beginning of file */
Xchar *LineStart,
X *LineEnd;
X{
X char OffStr[80];
X if (lFlag) {
X sprintf(OffStr,"%s\n",FileName);
X write(1,OffStr,strlen(OffStr));
X return;
X } /* if */
X if (FileName && !hFlag) {
X sprintf(OffStr,"%s: ",FileName);
X write(1,OffStr,strlen(OffStr));
X } /* if */
X if (nFlag) {
X sprintf(OffStr,"%d: ",OffSet);
X write(1,OffStr,strlen(OffStr));
X } /* if */
X write(1,LineStart,LineEnd-LineStart+1);
X if (*LineEnd != '\n') write (1,"\n",1);
X} /* PrintLine */
e-o-f
echo Extracting PutUsage.c
sed 's/^X//' > PutUsage.c << 'e-o-f'
X#include <stdio.h>
XPutUsage()
X{
X fprintf(stderr,
X "match: search for a given string or strings in a file or files\n");
X fprintf(stderr,
X "synopsis: match [option]* [string(s)] [file]*\n");
X fprintf(stderr,
X "options:\n");
X fprintf(stderr,
X "-c print only a count of matching lines \n");
X fprintf(stderr,
X "-e Take next argument as the pattern\n");
X fprintf(stderr,
X "-f PFile read patterns from a file PFile\n");
X fprintf(stderr,
X "-h Do not print file names\n");
X fprintf(stderr,
X "-l print a list of files containing the pattern(s) \n");
X fprintf(stderr,
X "-n print the character offset of the pattern(s) \n");
X fprintf(stderr,
X "-s silent mode. Return only success (0) or failure (1)\n");
X fprintf(stderr,
X "-x print only lines which match entirely \n");
X} /*PutUsage */
e-o-f
echo Extracting Search.s
sed 's/^X//' > Search.s << 'e-o-f'
XLL0:
X .data
X .text
X .align 1
X .globl _Search
X_Search:
X .word L20
X matchc 8(ap),*4(ap),16(ap),*12(ap)
X beql match
X movl 20(ap),r0
X movl r3,8(r0)
X movl $0,r0
X ret
Xmatch:
X movl 20(ap),r0
X movl r3,8(r0)
X movl $1,r0
X ret
X .set L20,0xf00
X .data
e-o-f
echo Extracting Makefile
sed 's/^X//' > Makefile << 'e-o-f'
XCFLAGS = -O
XSOURCES = Execute.c Extern.h\
X GetPatFile.c Global.c MakeDesc.c \
X match.h match.c MatchFound.c \
X MkDescVec.c MoveResidue.c PrintLine.c PutUsage.c Search.s
XOBJECTS = Execute.o \
X GetPatFile.o Global.o MakeDesc.o \
X match.o MatchFound.o \
X MkDescVec.o MoveResidue.o Search.o PrintLine.o PutUsage.o
XBASEFILES = $(SOURCES) Makefile README match.p
Xmatch: $(OBJECTS)
X cc -o match $(CFLAGS) $(OBJECTS)
Xshar:
X /usr/public/shar $(BASEFILES) >match.shar
Xinstall: match
X install -c match /usr/public/match
Xman: /usr/man/manp/match.p
X/usr/man/manp/match.p: match.p
X rm -f /usr/man/manp/match.p
X cp match.p /usr/man/manp/match.p
X man match > /dev/null
Xmatch.o: match.c match.h Extern.h
XPutUsage.o: PutUsage.c match.h
XSearch.o: Search.s
XExecute.o: Execute.c match.h
XMoveResidue.o: MoveResidue.c match.h Extern.h
XMatchFound.o: MatchFound.c match.h Extern.h
XPrintLine.o: PrintLine.c Extern.h
XMkDescVec.o: MkDescVec.c match.h
XGetPatFile.o: GetPatFile.c match.h
XMakeDesc.o: MakeDesc.c match.h
XGlobal.o: Global.c
Xlisting:
X print -i $(SOURCES) Makefile
Xclean:
X rm -f *.o match
e-o-f
echo Extracting README
sed 's/^X//' > README << 'e-o-f'
XMatch is a fast pattern matching utility, intended to be almost
Xidentical in functionality to fgrep (ugh!) but much faster. It uses
XMATCHC instruction on VAX computers. Apparently, it is faster than
XBM only on the VAX 780 - on the Microvax it is much slower than bm.
X
XThe files are:
XExecute.c: search a file for the patterns
XExtern.h: declarations of externs
XGetPatFile.c: read in patterns from a file and set up a vector of
X pattern descriptors
XGlobal.c: global variables (complement to Extern.h)
XMakeDesc.c: create a pattern descriptor for one pattern
XMakefile: you can figure this one out for yourself
XMatchFound.c: what to do when you actually FIND a pattern - print it,
X update flags, etc.
XMkDescVec.c: make a vector of pattern descriptors, given a string
X of newline-separated patterns
XMoveResidue.c: when you come to the end of the buffer, move the
X unsearched "residue" to the beginning and start again
XPrintLine.c: print the appropriate stuff after finding a match
XPutUsage.c: mini-man page.
XREADME: this file
XSearch.s: the guts. Calls MATCH
Xmatch.c: mainline. mostly interpreting the command line and tidying
X up at the end. Calls Execute for each file.
Xmatch.h: constants
Xmatch.p: man page
e-o-f
echo Extracting match.p
sed 's/^X//' > match.p << 'e-o-f'
X.TH MATCH PUBLIC "11 July 1985"
X.UC 4
X.SH NAME
Xmatch \- search a file for a string
X.SH SYNOPSIS
X.B /usr/public/match
X[ option ] ...
X[ strings ]
X[ file ]
X.SH DESCRIPTION
X.I Match
Xsearches the input
X.I files
X(standard input default) for lines matching a string.
XNormally, each line found is copied to the standard output.
XIt is blindingly fast.
X.I Match
Xstrings are fixed sequences of characters:
Xthere are no wildcards, repetitions, or other features
Xof regular expressions.
XMatch is also case sensitive.
XThe following options are recognized.
X.TP
X.B \-x
X(Exact) only lines matched in their entirety are printed
X.TP
X.B \-l
XThe names of files with matching lines are listed (once) separated by newlines.
X.TP
X.B \-c
XOnly a count of the number of matches
Xis printed
X.TP
X.B \-e "string"
XThe string is the next argument after the
X.B \-e
Xflag. This allows strings beginning with '-'.
X.TP
X.B \-h
XNo filenames are printed, even if multiple files are searched.
X.TP
X.B \-n
XEach line is preceded by the number
Xof characters from the beginning of the file
Xto the match.
X.TP
X.B \-s
XSilent mode. Nothing is printed (except error messages).
XThis is useful for checking the error status.
X.TP
X.BI \-f " file"
XThe string list
Xis taken from the
X.I file.
X.LP
XUnless the
X.B \-h
Xoption is specified
Xthe file name is shown if there is more than one input file.
XCare should be taken when using the characters $ * [ ^ | ( ) and \\ in the
X.I strings
X(listed on the command line)
Xas they are also meaningful to the Shell. It is safest to enclose the entire
X.I expression
Xargument in single quotes \' \'.
X.LP
X.I Match
Xsearches for lines that contain one of the (newline-separated)
X.I strings,
Xusing
Xthe VAX MATCHC instruction
XIt is far superior in terms of speed to the grep (egrep, fgrep)
Xfamily of pattern matchers, as well as bm(p) for fixed-pattern searching
Xon a VAX 780.
X.SH "SEE ALSO"
Xgrep(1), bm(p)
X.SH DIAGNOSTICS
XExit status is 0 if any matches are found,
X1 if none, 2 for syntax errors or inaccessible files.
X.SH AUTHOR
XPeter Bain (pdbain@wateng)
X.SH BUGS
XWorks slowly on VAXen other than the 780, and doesn't work at all
Xon other architectures.
X.LP
XOnly 100 patterns are allowed.
X.LP
XPatterns may not contain newlines.
X.LP
XIf a line (delimited by newlines, and the beginning and end of the file)
Xis longer than 8000 charcters (e.g. in a core dump),
Xit will not be completely printed.
X.LP
XIf multiple patterns are specified, the order of the ouput lines is not
Xnecessarily the same as the order of the input lines.
X.LP
XA line will be printed once for each different string on that line.
X.LP
XThe algorithm cannot count lines.
X.LP
XThe
X.B -n
Xand
X.B -c
Xwork differently from fgrep.
X.LP
XThe
X.B -v,
X.B -i,
Xand
X.B -b
Xare not available.
e-o-f
exit 0
--
- peter bain
...!{allegra|decvax|clyde|ihnp4 }!watmath!wateng!pdbain
hard mail: CCNG, CPH-2369A, University of Waterloo,
Waterloo, Ont. Canada N2M 5G4
telephone: (519) 885-1211 x2810