d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
84c4de2d47
freebsd-nq/usr.bin/sort/file.c
Gabor Kovesdan 5ca724dc59 - Only use multi-threading for large files 
- Do not use mmap() by default; it can be enabled by --mmap
- Add some minor optimizations for -u
- Update manual page according to the changes

Submitted by:	Oleg Moskalenko <oleg.moskalenko@citrix.com>
2012-05-25 09:30:16 +00:00

1630 lines
31 KiB
C
Raw Blame History

/*-
* Copyright (C) 2009 Gabor Kovesdan <gabor@FreeBSD.org>
* Copyright (C) 2012 Oleg Moskalenko <oleg.moskalenko@citrix.com>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/mman.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/queue.h>
#include <err.h>
#include <fcntl.h>
#if defined(SORT_THREADS)
#include <pthread.h>
#endif
#include <semaphore.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <wchar.h>
#include <wctype.h>
#include "coll.h"
#include "file.h"
#include "radixsort.h"
unsigned long long free_memory = 1000000;
unsigned long long available_free_memory = 1000000;
bool use_mmap;
const char *tmpdir = "/var/tmp";
const char *compress_program;
size_t max_open_files = 16;
/*
* How much space we read from file at once
*/
#define READ_CHUNK (4096)
/*
* File reader structure
*/
struct file_reader
{
struct reader_buffer rb;
FILE *file;
char *fname;
unsigned char *buffer;
unsigned char *mmapaddr;
unsigned char *mmapptr;
size_t bsz;
size_t cbsz;
size_t mmapsize;
size_t strbeg;
int fd;
char elsymb;
};
/*
* Structure to be used in file merge process.
*/
struct file_header
{
struct file_reader *fr;
struct sort_list_item *si; /* current top line */
size_t file_pos;
};
/*
* List elements of "cleanable" files list.
*/
struct CLEANABLE_FILE
{
char *fn;
LIST_ENTRY(CLEANABLE_FILE) files;
};
/*
* List header of "cleanable" files list.
*/
static LIST_HEAD(CLEANABLE_FILES,CLEANABLE_FILE) tmp_files;
/*
* Semaphore to protect the tmp file list.
* We use semaphore here because it is signal-safe, according to POSIX.
* And semaphore does not require pthread library.
*/
static sem_t tmp_files_sem;
static void mt_sort(struct sort_list *list,
int (*sort_func)(void *, size_t, size_t,
int (*)(const void *, const void *)), const char* fn);
/*
* Init tmp files list
*/
void
init_tmp_files(void)
{
LIST_INIT(&tmp_files);
sem_init(&tmp_files_sem, 0, 1);
}
/*
* Save name of a tmp file for signal cleanup
*/
void
tmp_file_atexit(const char *tmp_file)
{
if (tmp_file) {
sem_wait(&tmp_files_sem);
struct CLEANABLE_FILE *item =
sort_malloc(sizeof(struct CLEANABLE_FILE));
item->fn = sort_strdup(tmp_file);
LIST_INSERT_HEAD(&tmp_files, item, files);
sem_post(&tmp_files_sem);
}
}
/*
* Clear tmp files
*/
void
clear_tmp_files(void)
{
struct CLEANABLE_FILE *item;
sem_wait(&tmp_files_sem);
LIST_FOREACH(item,&tmp_files,files) {
if ((item) && (item->fn))
unlink(item->fn);
}
sem_post(&tmp_files_sem);
}
/*
* Check whether a file is a temporary file
*/
static bool
file_is_tmp(const char* fn)
{
struct CLEANABLE_FILE *item;
bool ret = false;
if (fn) {
sem_wait(&tmp_files_sem);
LIST_FOREACH(item,&tmp_files,files) {
if ((item) && (item->fn))
if (strcmp(item->fn, fn) == 0) {
ret = true;
break;
}
}
sem_post(&tmp_files_sem);
}
return (ret);
}
/*
* Read zero-terminated line from a file
*/
char *
read_file0_line(struct file0_reader *f0r)
{
size_t pos = 0;
int c;
if ((f0r->f == NULL) || feof(f0r->f))
return (NULL);
if (f0r->current_line && f0r->current_sz > 0)
f0r->current_line[0] = 0;
while (!feof(f0r->f)) {
c = fgetc(f0r->f);
if (feof(f0r->f) || (c == -1))
break;
if ((pos + 1) >= f0r->current_sz) {
size_t newsz = (f0r->current_sz + 2) * 2;
f0r->current_line = sort_realloc(f0r->current_line,
newsz);
f0r->current_sz = newsz;
}
f0r->current_line[pos] = (char)c;
if (c == 0)
break;
else
f0r->current_line[pos + 1] = 0;
++pos;
}
return f0r->current_line;
}
/*
* Generate new temporary file name
*/
char *
new_tmp_file_name(void)
{
static unsigned int tfcounter = 0;
static const char *fn = ".bsdsort.";
char *ret;
size_t sz;
sz = strlen(tmpdir) + 1 + strlen(fn) + 32 + 1;
ret = sort_malloc(sz);
sprintf(ret, "%s/%s%d.%u", tmpdir, fn, (int) getpid(), tfcounter++);
tmp_file_atexit(ret);
return (ret);
}
/*
* Initialize file list
*/
void
file_list_init(struct file_list *fl, bool tmp)
{
if (fl) {
fl->count = 0;
fl->sz = 0;
fl->fns = NULL;
fl->tmp = tmp;
}
}
/*
* Add a file name to the list
*/
void
file_list_add(struct file_list *fl, char *fn, bool allocate)
{
if (fl && fn) {
if (fl->count >= fl->sz || (fl->fns == NULL)) {
fl->sz = (fl->sz) * 2 + 1;
fl->fns = sort_realloc(fl->fns, fl->sz *
sizeof(char *));
}
fl->fns[fl->count] = allocate ? sort_strdup(fn) : fn;
fl->count += 1;
}
}
/*
* Populate file list from array of file names
*/
void
file_list_populate(struct file_list *fl, int argc, char **argv, bool allocate)
{
if (fl && argv) {
int i;
for (i = 0; i < argc; i++)
file_list_add(fl, argv[i], allocate);
}
}
/*
* Clean file list data and delete the files,
* if this is a list of temporary files
*/
void
file_list_clean(struct file_list *fl)
{
if (fl) {
if (fl->fns) {
int i;
for (i = 0; i < fl->count; i++) {
if (fl->fns[i]) {
if (fl->tmp)
unlink(fl->fns[i]);
sort_free(fl->fns[i]);
fl->fns[i] = 0;
}
}
sort_free(fl->fns);
fl->fns = NULL;
}
fl->sz = 0;
fl->count = 0;
fl->tmp = false;
}
}
/*
* Init sort list
*/
void
sort_list_init(struct sort_list *l)
{
if (l) {
l->count = 0;
l->size = 0;
l->memsize = sizeof(struct sort_list);
l->list = NULL;
}
}
/*
* Add string to sort list
*/
void
sort_list_add(struct sort_list *l, struct bwstring *str)
{
if (l && str) {
size_t indx = l->count;
if ((l->list == NULL) || (indx >= l->size)) {
int newsize = (l->size + 1) + 1024;
l->list = sort_realloc(l->list,
sizeof(struct sort_list_item*) * newsize);
l->memsize += (newsize - l->size) *
sizeof(struct sort_list_item*);
l->size = newsize;
}
l->list[indx] = sort_list_item_alloc();
sort_list_item_set(l->list[indx], str);
l->memsize += sort_list_item_size(l->list[indx]);
l->count += 1;
}
}
/*
* Clean sort list data
*/
void
sort_list_clean(struct sort_list *l)
{
if (l) {
if (l->list) {
size_t i;
for (i = 0; i < l->count; i++) {
struct sort_list_item *item;
item = l->list[i];
if (item) {
sort_list_item_clean(item);
sort_free(item);
l->list[i] = NULL;
}
}
sort_free(l->list);
l->list = NULL;
}
l->count = 0;
l->size = 0;
l->memsize = sizeof(struct sort_list);
}
}
/*
* Write sort list to file
*/
void
sort_list_dump(struct sort_list *l, const char *fn)
{
if (l && fn) {
FILE *f;
f = openfile(fn, "w");
if (f == NULL)
err(2, NULL);
if (l->list) {
size_t i;
if (!(sort_opts_vals.uflag)) {
for (i = 0; i < l->count; ++i)
bwsfwrite(l->list[i]->str, f,
sort_opts_vals.zflag);
} else {
struct sort_list_item *last_printed_item = NULL;
struct sort_list_item *item;
for (i = 0; i < l->count; ++i) {
item = l->list[i];
if ((last_printed_item == NULL) ||
list_coll(&last_printed_item, &item)) {
bwsfwrite(item->str, f, sort_opts_vals.zflag);
last_printed_item = item;
}
}
}
}
closefile(f, fn);
}
}
/*
* Checks if the given file is sorted. Stops at the first disorder,
* prints the disordered line and returns 1.
*/
int
check(const char *fn)
{
struct bwstring *s1, *s2, *s1disorder, *s2disorder;
struct file_reader *fr;
struct keys_array *ka1, *ka2;
int res, pos, posdisorder;
s1 = s2 = s1disorder = s2disorder = NULL;
ka1 = ka2 = NULL;
fr = file_reader_init(fn);
res = 0;
pos = 1;
posdisorder = 1;
if (fr == NULL) {
err(2, NULL);
goto end;
}
s1 = file_reader_readline(fr);
if (s1 == NULL)
goto end;
ka1 = keys_array_alloc();
preproc(s1, ka1);
s2 = file_reader_readline(fr);
if (s2 == NULL)
goto end;
ka2 = keys_array_alloc();
preproc(s2, ka2);
for (;;) {
if (debug_sort) {
bwsprintf(stdout, s2, "s1=<", ">");
bwsprintf(stdout, s1, "s2=<", ">");
}
int cmp = key_coll(ka2, ka1, 0);
if (debug_sort)
printf("; cmp1=%d", cmp);
if (!cmp && sort_opts_vals.complex_sort &&
!(sort_opts_vals.uflag) && !(sort_opts_vals.sflag)) {
cmp = top_level_str_coll(s2, s1);
if (debug_sort)
printf("; cmp2=%d", cmp);
}
if (debug_sort)
printf("\n");
if ((sort_opts_vals.uflag && (cmp <= 0)) || (cmp < 0)) {
if (!(sort_opts_vals.csilentflag)) {
s2disorder = bwsdup(s2);
posdisorder = pos;
if (debug_sort)
s1disorder = bwsdup(s1);
}
res = 1;
goto end;
}
pos++;
clean_keys_array(s1, ka1);
sort_free(ka1);
ka1 = ka2;
ka2 = NULL;
bwsfree(s1);
s1 = s2;
s2 = file_reader_readline(fr);
if (s2 == NULL)
goto end;
ka2 = keys_array_alloc();
preproc(s2, ka2);
}
end:
if (ka1) {
clean_keys_array(s1, ka1);
sort_free(ka1);
}
if (s1)
bwsfree(s1);
if (ka2) {
clean_keys_array(s2, ka2);
sort_free(ka2);
}
if (s2)
bwsfree(s2);
if ((fn == NULL) || (*fn == 0) || (strcmp(fn, "-") == 0)) {
for (;;) {
s2 = file_reader_readline(fr);
if (s2 == NULL)
break;
bwsfree(s2);
}
}
file_reader_free(fr);
if (s2disorder) {
bws_disorder_warnx(s2disorder, fn, posdisorder);
if (s1disorder) {
bws_disorder_warnx(s1disorder, fn, posdisorder);
if (s1disorder != s2disorder)
bwsfree(s1disorder);
}
bwsfree(s2disorder);
s1disorder = NULL;
s2disorder = NULL;
}
if (res)
exit(res);
return (0);
}
/*
* Opens a file. If the given filename is "-", stdout will be
* opened.
*/
FILE *
openfile(const char *fn, const char *mode)
{
FILE *file;
if (strcmp(fn, "-") == 0) {
return ((mode && mode[0] == 'r') ? stdin : stdout);
} else {
mode_t orig_file_mask = 0;
int is_tmp = file_is_tmp(fn);
if (is_tmp && (mode[0] == 'w'))
orig_file_mask = umask(S_IWGRP | S_IWOTH |
S_IRGRP | S_IROTH);
if (is_tmp && (compress_program != NULL)) {
char *cmd;
size_t cmdsz;
cmdsz = strlen(fn) + 128;
cmd = sort_malloc(cmdsz);
fflush(stdout);
if (mode[0] == 'r')
snprintf(cmd, cmdsz - 1, "cat %s | %s -d",
fn, compress_program);
else if (mode[0] == 'w')
snprintf(cmd, cmdsz - 1, "%s > %s",
compress_program, fn);
else
err(2, "%s", getstr(7));
if ((file = popen(cmd, mode)) == NULL)
err(2, NULL);
sort_free(cmd);
} else
if ((file = fopen(fn, mode)) == NULL)
err(2, NULL);
if (is_tmp && (mode[0] == 'w'))
umask(orig_file_mask);
}
return (file);
}
/*
* Close file
*/
void
closefile(FILE *f, const char *fn)
{
if (f == NULL) {
;
} else if (f == stdin) {
;
} else if (f == stdout) {
fflush(f);
} else {
if (file_is_tmp(fn) && compress_program != NULL) {
if(pclose(f)<0)
err(2,NULL);
} else
fclose(f);
}
}
/*
* Reads a file into the internal buffer.
*/
struct file_reader *
file_reader_init(const char *fsrc)
{
struct file_reader *ret;
if (fsrc == NULL)
fsrc = "-";
ret = sort_malloc(sizeof(struct file_reader));
memset(ret, 0, sizeof(struct file_reader));
ret->elsymb = '\n';
if (sort_opts_vals.zflag)
ret->elsymb = 0;
ret->fname = sort_strdup(fsrc);
if (strcmp(fsrc, "-") && (compress_program == NULL) && use_mmap) {
do {
struct stat stat_buf;
void *addr;
size_t sz = 0;
int fd, flags;
flags = MAP_NOCORE | MAP_NOSYNC;
addr = MAP_FAILED;
fd = open(fsrc, O_RDONLY);
if (fd < 0)
err(2, NULL);
if (fstat(fd, &stat_buf) < 0) {
close(fd);
break;
}
sz = stat_buf.st_size;
#if defined(MAP_PREFAULT_READ)
flags |= MAP_PREFAULT_READ;
#endif
addr = mmap(NULL, sz, PROT_READ, flags, fd, 0);
if (addr == MAP_FAILED) {
close(fd);
break;
}
ret->fd = fd;
ret->mmapaddr = addr;
ret->mmapsize = sz;
ret->mmapptr = ret->mmapaddr;
} while (0);
}
if (ret->mmapaddr == NULL) {
ret->file = openfile(fsrc, "r");
if (ret->file == NULL)
err(2, NULL);
if (strcmp(fsrc, "-")) {
ret->cbsz = READ_CHUNK;
ret->buffer = sort_malloc(ret->cbsz);
ret->bsz = 0;
ret->strbeg = 0;
ret->bsz = fread(ret->buffer, 1, ret->cbsz, ret->file);
if (ret->bsz == 0) {
if (ferror(ret->file))
err(2, NULL);
}
}
}
return (ret);
}
struct bwstring *
file_reader_readline(struct file_reader *fr)
{
struct bwstring *ret = NULL;
if (fr->mmapaddr) {
unsigned char *mmapend;
mmapend = fr->mmapaddr + fr->mmapsize;
if (fr->mmapptr >= mmapend)
return (NULL);
else {
unsigned char *strend;
size_t sz;
sz = mmapend - fr->mmapptr;
strend = memchr(fr->mmapptr, fr->elsymb, sz);
if (strend == NULL) {
ret = bwscsbdup(fr->mmapptr, sz);
fr->mmapptr = mmapend;
} else {
ret = bwscsbdup(fr->mmapptr, strend -
fr->mmapptr);
fr->mmapptr = strend + 1;
}
}
} else if (fr->file != stdin) {
unsigned char *strend;
size_t bsz1, remsz, search_start;
search_start = 0;
remsz = 0;
strend = NULL;
if (fr->bsz > fr->strbeg)
remsz = fr->bsz - fr->strbeg;
/* line read cycle */
for (;;) {
if (remsz > search_start)
strend = memchr(fr->buffer + fr->strbeg +
search_start, fr->elsymb, remsz -
search_start);
else
strend = NULL;
if (strend)
break;
if (feof(fr->file))
break;
if (fr->bsz != fr->cbsz)
/* NOTREACHED */
err(2, "File read software error 1");
if (remsz > (READ_CHUNK >> 1)) {
search_start = fr->cbsz - fr->strbeg;
fr->cbsz += READ_CHUNK;
fr->buffer = sort_realloc(fr->buffer,
fr->cbsz);
bsz1 = fread(fr->buffer + fr->bsz, 1,
READ_CHUNK, fr->file);
if (bsz1 == 0) {
if (ferror(fr->file))
err(2, NULL);
break;
}
fr->bsz += bsz1;
remsz += bsz1;
} else {
if (remsz > 0 && fr->strbeg>0)
bcopy(fr->buffer + fr->strbeg,
fr->buffer, remsz);
fr->strbeg = 0;
search_start = remsz;
bsz1 = fread(fr->buffer + remsz, 1,
fr->cbsz - remsz, fr->file);
if (bsz1 == 0) {
if (ferror(fr->file))
err(2, NULL);
break;
}
fr->bsz = remsz + bsz1;
remsz = fr->bsz;
}
}
if (strend == NULL)
strend = fr->buffer + fr->bsz;
if ((fr->buffer + fr->strbeg <= strend) &&
(fr->strbeg < fr->bsz) && (remsz>0))
ret = bwscsbdup(fr->buffer + fr->strbeg, strend -
fr->buffer - fr->strbeg);
fr->strbeg = (strend - fr->buffer) + 1;
} else {
size_t len = 0;
ret = bwsfgetln(fr->file, &len, sort_opts_vals.zflag,
&(fr->rb));
}
return (ret);
}
static void
file_reader_clean(struct file_reader *fr)
{
if (fr) {
if (fr->mmapaddr)
munmap(fr->mmapaddr, fr->mmapsize);
if (fr->fd)
close(fr->fd);
if (fr->buffer)
sort_free(fr->buffer);
if (fr->file)
if (fr->file != stdin)
closefile(fr->file, fr->fname);
if(fr->fname)
sort_free(fr->fname);
memset(fr, 0, sizeof(struct file_reader));
}
}
void
file_reader_free(struct file_reader *fr)
{
if (fr) {
file_reader_clean(fr);
sort_free(fr);
}
}
int
procfile(const char *fsrc, struct sort_list *list, struct file_list *fl)
{
struct file_reader *fr;
fr = file_reader_init(fsrc);
if (fr == NULL)
err(2, NULL);
/* file browse cycle */
for (;;) {
struct bwstring *bws;
bws = file_reader_readline(fr);
if (bws == NULL)
break;
sort_list_add(list, bws);
if (list->memsize >= available_free_memory) {
char *fn;
fn = new_tmp_file_name();
sort_list_to_file(list, fn);
file_list_add(fl, fn, false);
sort_list_clean(list);
}
}
file_reader_free(fr);
return (0);
}
/*
* Compare file headers. Files with EOF always go to the end of the list.
*/
static int
file_header_cmp(struct file_header *f1, struct file_header *f2)
{
if (f1 == f2)
return (0);
else {
if (f1->fr == NULL) {
return ((f2->fr == NULL) ? 0 : +1);
} else if (f2->fr == NULL)
return (-1);
else {
int ret;
ret = list_coll(&(f1->si), &(f2->si));
if (!ret)
return ((f1->file_pos < f2->file_pos) ? -1 : +1);
return (ret);
}
}
}
/*
* Allocate and init file header structure
*/
static void
file_header_init(struct file_header **fh, const char *fn, size_t file_pos)
{
if (fh && fn) {
struct bwstring *line;
*fh = sort_malloc(sizeof(struct file_header));
(*fh)->file_pos = file_pos;
(*fh)->fr = file_reader_init(fn);
if ((*fh)->fr == NULL) {
perror(fn);
err(2, "%s", getstr(8));
}
line = file_reader_readline((*fh)->fr);
if (line == NULL) {
file_reader_free((*fh)->fr);
(*fh)->fr = NULL;
(*fh)->si = NULL;
} else {
(*fh)->si = sort_list_item_alloc();
sort_list_item_set((*fh)->si, line);
}
}
}
/*
* Close file
*/
static void
file_header_close(struct file_header **fh)
{
if (fh && *fh) {
if ((*fh)->fr) {
file_reader_free((*fh)->fr);
(*fh)->fr = NULL;
}
if ((*fh)->si) {
sort_list_item_clean((*fh)->si);
sort_free((*fh)->si);
(*fh)->si = NULL;
}
sort_free(*fh);
*fh = NULL;
}
}
/*
* Swap two array elements
*/
static void
file_header_swap(struct file_header **fh, int i1, int i2)
{
struct file_header *tmp;
tmp = fh[i1];
fh[i1] = fh[i2];
fh[i2] = tmp;
}
/* heap algorithm ==>> */
/*
* See heap sort algorithm
* "Raises" last element to its right place
*/
static void
file_header_heap_swim(struct file_header **fh, int indx)
{
if (indx > 0) {
int parent_index;
parent_index = (indx - 1) >> 1;
if (file_header_cmp(fh[indx], fh[parent_index]) < 0) {
/* swap child and parent and continue */
file_header_swap(fh, indx, parent_index);
file_header_heap_swim(fh, parent_index);
}
}
}
/*
* Sink the top element to its correct position
*/
static void
file_header_heap_sink(struct file_header **fh, int indx, int size)
{
int left_child_index;
int right_child_index;
left_child_index = indx + indx + 1;
right_child_index = left_child_index + 1;
if (left_child_index < size) {
int min_child_index;
min_child_index = left_child_index;
if ((right_child_index < size) &&
(file_header_cmp(fh[left_child_index],
fh[right_child_index]) > 0))
min_child_index = right_child_index;
if (file_header_cmp(fh[indx], fh[min_child_index]) > 0) {
file_header_swap(fh, indx, min_child_index);
file_header_heap_sink(fh, min_child_index, size);
}
}
}
/* <<== heap algorithm */
/*
* Adds element to the "left" end
*/
static void
file_header_list_rearrange_from_header(struct file_header **fh, int size)
{
file_header_heap_sink(fh, 0, size);
}
/*
* Adds element to the "right" end
*/
static void
file_header_list_push(struct file_header *f, struct file_header **fh, int size)
{
fh[size++] = f;
file_header_heap_swim(fh, size - 1);
}
struct last_printed
{
struct bwstring *str;
};
/*
* Prints the current line of the file
*/
static void
file_header_print(struct file_header *fh, FILE *f_out, struct last_printed *lp)
{
if (fh && fh->fr && f_out && fh->si && fh->si->str) {
if (sort_opts_vals.uflag) {
if ((lp->str == NULL) || (str_list_coll(lp->str, &(fh->si)))) {
bwsfwrite(fh->si->str, f_out, sort_opts_vals.zflag);
if (lp->str)
bwsfree(lp->str);
lp->str = bwsdup(fh->si->str);
}
} else
bwsfwrite(fh->si->str, f_out, sort_opts_vals.zflag);
}
}
/*
* Read next line
*/
static void
file_header_read_next(struct file_header *fh)
{
if (fh && fh->fr) {
struct bwstring *tmp;
tmp = file_reader_readline(fh->fr);
if (tmp == NULL) {
file_reader_free(fh->fr);
fh->fr = NULL;
if (fh->si) {
sort_list_item_clean(fh->si);
sort_free(fh->si);
fh->si = NULL;
}
} else {
if (fh->si == NULL)
fh->si = sort_list_item_alloc();
sort_list_item_set(fh->si, tmp);
}
}
}
/*
* Merge array of "files headers"
*/
static void
file_headers_merge(int fnum, struct file_header **fh, FILE *f_out)
{
struct last_printed lp;
int i;
memset(&lp, 0, sizeof(lp));
/*
* construct the initial sort structure
*/
for (i = 0; i < fnum; i++)
file_header_list_push(fh[i], fh, i);
while (fh[0]->fr) { /* unfinished files are always in front */
/* output the smallest line: */
file_header_print(fh[0], f_out, &lp);
/* read a new line, if possible: */
file_header_read_next(fh[0]);
/* re-arrange the list: */
file_header_list_rearrange_from_header(fh, fnum);
}
if (lp.str)
bwsfree(lp.str);
}
/*
* Merges the given files into the output file, which can be
* stdout.
*/
static void
merge_files_array(int argc, char **argv, const char *fn_out)
{
if (argv && fn_out) {
struct file_header **fh;
FILE *f_out;
int i;
f_out = openfile(fn_out, "w");
if (f_out == NULL)
err(2, NULL);
fh = sort_malloc((argc + 1) * sizeof(struct file_header *));
for (i = 0; i < argc; i++)
file_header_init(fh + i, argv[i], (size_t) i);
file_headers_merge(argc, fh, f_out);
for (i = 0; i < argc; i++)
file_header_close(fh + i);
sort_free(fh);
closefile(f_out, fn_out);
}
}
/*
* Shrinks the file list until its size smaller than max number of opened files
*/
static int
shrink_file_list(struct file_list *fl)
{
if ((fl == NULL) || (size_t) (fl->count) < max_open_files)
return (0);
else {
struct file_list new_fl;
int indx = 0;
file_list_init(&new_fl, true);
while (indx < fl->count) {
char *fnew;
int num;
num = fl->count - indx;
fnew = new_tmp_file_name();
if ((size_t) num >= max_open_files)
num = max_open_files - 1;
merge_files_array(num, fl->fns + indx, fnew);
if (fl->tmp) {
int i;
for (i = 0; i < num; i++)
unlink(fl->fns[indx + i]);
}
file_list_add(&new_fl, fnew, false);
indx += num;
}
fl->tmp = false; /* already taken care of */
file_list_clean(fl);
fl->count = new_fl.count;
fl->fns = new_fl.fns;
fl->sz = new_fl.sz;
fl->tmp = new_fl.tmp;
return (1);
}
}
/*
* Merge list of files
*/
void
merge_files(struct file_list *fl, const char *fn_out)
{
if (fl && fn_out) {
while (shrink_file_list(fl));
merge_files_array(fl->count, fl->fns, fn_out);
}
}
static const char *
get_sort_method_name(int sm)
{
if (sm == SORT_MERGESORT)
return "mergesort";
else if (sort_opts_vals.sort_method == SORT_RADIXSORT)
return "radixsort";
else if (sort_opts_vals.sort_method == SORT_HEAPSORT)
return "heapsort";
else
return "quicksort";
}
/*
* Wrapper for qsort
*/
static int sort_qsort(void *list, size_t count, size_t elem_size,
int (*cmp_func)(const void *, const void *))
{
qsort(list, count, elem_size, cmp_func);
return (0);
}
/*
* Sort list of lines and writes it to the file
*/
void
sort_list_to_file(struct sort_list *list, const char *outfile)
{
struct sort_mods *sm = &(keys[0].sm);
if (!(sm->Mflag) && !(sm->Rflag) && !(sm->Vflag) && !(sm->Vflag) &&
!(sm->gflag) && !(sm->hflag) && !(sm->nflag)) {
if ((sort_opts_vals.sort_method == SORT_DEFAULT) && byte_sort)
sort_opts_vals.sort_method = SORT_RADIXSORT;
} else if (sort_opts_vals.sort_method == SORT_RADIXSORT)
err(2, "%s", getstr(9));
/*
* to handle stable sort and the unique cases in the
* right order, we need stable basic algorithm
*/
if (sort_opts_vals.sflag) {
switch (sort_opts_vals.sort_method){
case SORT_MERGESORT:
break;
case SORT_RADIXSORT:
break;
case SORT_DEFAULT:
sort_opts_vals.sort_method = SORT_MERGESORT;
break;
default:
errx(2, "%s", getstr(10));
};
}
if (sort_opts_vals.sort_method == SORT_DEFAULT)
sort_opts_vals.sort_method = SORT_QSORT;
if (debug_sort)
printf("sort_method=%s\n",
get_sort_method_name(sort_opts_vals.sort_method));
switch (sort_opts_vals.sort_method){
case SORT_RADIXSORT:
rxsort(list->list, list->count);
sort_list_dump(list, outfile);
break;
case SORT_MERGESORT:
mt_sort(list, mergesort, outfile);
break;
case SORT_HEAPSORT:
mt_sort(list, heapsort, outfile);
break;
default:
mt_sort(list, sort_qsort, outfile);
break;
}
}
/******************* MT SORT ************************/
#if defined(SORT_THREADS)
/* semaphore to count threads */
static sem_t mtsem;
/* current system sort function */
static int (*g_sort_func)(void *, size_t, size_t,
int(*)(const void *, const void *));
/*
* Sort cycle thread (in multi-threaded mode)
*/
static void*
mt_sort_thread(void* arg)
{
struct sort_list *list = arg;
g_sort_func(list->list, list->count, sizeof(struct sort_list_item *),
(int(*)(const void *, const void *)) list_coll);
sem_post(&mtsem);
return (arg);
}
/*
* Compare sub-lists. Empty sub-lists always go to the end of the list.
*/
static int
sub_list_cmp(struct sort_list *l1, struct sort_list *l2)
{
if (l1 == l2)
return (0);
else {
if (l1->count == 0) {
return ((l2->count == 0) ? 0 : +1);
} else if (l2->count == 0) {
return (-1);
} else {
int ret;
ret = list_coll(&(l1->list[0]), &(l2->list[0]));
if (!ret)
return ((l1->sub_list_pos < l2->sub_list_pos) ?
-1 : +1);
return (ret);
}
}
}
/*
* Swap two array elements
*/
static void
sub_list_swap(struct sort_list **sl, int i1, int i2)
{
struct sort_list *tmp;
tmp = sl[i1];
sl[i1] = sl[i2];
sl[i2] = tmp;
}
/* heap algorithm ==>> */
/*
* See heap sort algorithm
* "Raises" last element to its right place
*/
static void
sub_list_swim(struct sort_list **sl, int indx)
{
if (indx > 0) {
int parent_index;
parent_index = (indx - 1) >> 1;
if (sub_list_cmp(sl[indx], sl[parent_index]) < 0) {
/* swap child and parent and continue */
sub_list_swap(sl, indx, parent_index);
sub_list_swim(sl, parent_index);
}
}
}
/*
* Sink the top element to its correct position
*/
static void
sub_list_sink(struct sort_list **sl, int indx, int size)
{
int left_child_index;
int right_child_index;
left_child_index = indx + indx + 1;
right_child_index = left_child_index + 1;
if (left_child_index < size) {
int min_child_index;
min_child_index = left_child_index;
if ((right_child_index < size) &&
(sub_list_cmp(sl[left_child_index],
sl[right_child_index]) > 0))
min_child_index = right_child_index;
if (sub_list_cmp(sl[indx], sl[min_child_index]) > 0) {
sub_list_swap(sl, indx, min_child_index);
sub_list_sink(sl, min_child_index, size);
}
}
}
/* <<== heap algorithm */
/*
* Adds element to the "right" end
*/
static void
sub_list_push(struct sort_list *s, struct sort_list **sl, int size)
{
sl[size++] = s;
sub_list_swim(sl, size - 1);
}
struct last_printed_item
{
struct sort_list_item *item;
};
/*
* Prints the current line of the file
*/
static void
sub_list_header_print(struct sort_list *sl, FILE *f_out,
struct last_printed_item *lp)
{
if (sl && sl->count && f_out && sl->list[0]->str) {
if (sort_opts_vals.uflag) {
if ((lp->item == NULL) || (list_coll(&(lp->item),
&(sl->list[0])))) {
bwsfwrite(sl->list[0]->str, f_out,
sort_opts_vals.zflag);
lp->item = sl->list[0];
}
} else
bwsfwrite(sl->list[0]->str, f_out,
sort_opts_vals.zflag);
}
}
/*
* Read next line
*/
static void
sub_list_next(struct sort_list *sl)
{
if (sl && sl->count) {
sl->list += 1;
sl->count -= 1;
}
}
/*
* Merge sub-lists to a file
*/
static void
merge_sub_lists(struct sort_list **sl, size_t n, FILE* f_out)
{
struct last_printed_item lp;
size_t i;
memset(&lp,0,sizeof(lp));
/* construct the initial list: */
for (i = 0; i < n; i++)
sub_list_push(sl[i], sl, i);
while (sl[0]->count) { /* unfinished lists are always in front */
/* output the smallest line: */
sub_list_header_print(sl[0], f_out, &lp);
/* move to a new line, if possible: */
sub_list_next(sl[0]);
/* re-arrange the list: */
sub_list_sink(sl, 0, n);
}
}
/*
* Merge sub-lists to a file
*/
static void
merge_list_parts(struct sort_list **parts, size_t n, const char *fn)
{
FILE* f_out;
f_out = openfile(fn,"w");
merge_sub_lists(parts, n, f_out);
closefile(f_out, fn);
}
#endif /* defined(SORT_THREADS) */
/*
* Multi-threaded sort algorithm "driver"
*/
static void
mt_sort(struct sort_list *list,
int(*sort_func)(void *, size_t, size_t, int(*)(const void *, const void *)),
const char* fn)
{
#if defined(SORT_THREADS)
if (nthreads < 2 || list->count < MT_SORT_THRESHOLD) {
size_t nthreads_save = nthreads;
nthreads = 1;
#endif
/* if single thread or small data, do simple sort */
sort_func(list->list, list->count,
sizeof(struct sort_list_item *),
(int(*)(const void *, const void *)) list_coll);
sort_list_dump(list, fn);
#if defined(SORT_THREADS)
nthreads = nthreads_save;
} else {
/* multi-threaded sort */
struct sort_list **parts;
size_t avgsize, cstart, i;
/* array of sub-lists */
parts = sort_malloc(sizeof(struct sort_list*) * nthreads);
cstart = 0;
avgsize = list->count / nthreads;
/* set global system sort function */
g_sort_func = sort_func;
/* set sublists */
for (i = 0; i < nthreads; ++i) {
size_t sz = 0;
parts[i] = sort_malloc(sizeof(struct sort_list));
parts[i]->list = list->list + cstart;
parts[i]->memsize = 0;
parts[i]->sub_list_pos = i;
sz = (i == nthreads - 1) ? list->count - cstart :
avgsize;
parts[i]->count = sz;
parts[i]->size = parts[i]->count;
cstart += sz;
}
/* init threads counting semaphore */
sem_init(&mtsem, 0, 0);
/* start threads */
for (i = 0; i < nthreads; ++i) {
pthread_t pth;
pthread_attr_t attr;
pthread_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_DETACHED);
for (;;) {
int res = pthread_create(&pth, &attr,
mt_sort_thread, parts[i]);
if (res >= 0)
break;
if (errno == EAGAIN) {
pthread_yield();
continue;
}
err(2, NULL);
}
pthread_attr_destroy(&attr);
}
/* wait for threads completion */
for (i = 0; i < nthreads; ++i) {
sem_wait(&mtsem);
}
/* destroy the semaphore - we do not need it anymore */
sem_destroy(&mtsem);
/* merge sorted sub-lists to the file */
merge_list_parts(parts, nthreads, fn);
/* free sub-lists data */
for (i = 0; i < nthreads; ++i) {
sort_free(parts[i]);
}
sort_free(parts);
}
#endif /* defined(SORT_THREADS) */
}
Reference in New Issue View Git Blame Copy Permalink