/* //////////////////////////////////////////////////////////////////////////////////////
* interfaces
*/
tb_timer_ref_t tb_timer_init(tb_size_t maxn, tb_bool_t ctime)
{
// done
tb_bool_t ok = tb_false;
tb_timer_impl_t* impl = tb_null;
do
{
// make timer
impl = tb_malloc0_type(tb_timer_impl_t);
tb_assert_and_check_break(impl);
// init func
tb_item_func_t func = tb_item_func_ptr(tb_null, tb_null); func.comp = tb_timer_comp_by_when;
// init timer
impl->maxn = tb_max(maxn, 16);
impl->ctime = ctime;
// init lock
if (!tb_spinlock_init(&impl->lock)) break;
// init pool
impl->pool = tb_fixed_pool_init(tb_null, (maxn >> 4) + 16, sizeof(tb_timer_task_impl_t), tb_null, tb_null, tb_null);
tb_assert_and_check_break(impl->pool);
// init heap
impl->heap = tb_heap_init((maxn >> 2) + 16, func);
tb_assert_and_check_break(impl->heap);
// register lock profiler
#ifdef TB_LOCK_PROFILER_ENABLE
tb_lock_profiler_register(tb_lock_profiler(), (tb_pointer_t)&impl->lock, TB_TRACE_MODULE_NAME);
#endif
// ok
ok = tb_true;
} while (0);
// failed?
if (!ok)
{
// exit it
if (impl) tb_timer_exit((tb_timer_ref_t)impl);
impl = tb_null;
}
// ok?
return (tb_timer_ref_t)impl;
}
tb_char_t const* tb_regex_replace(tb_regex_ref_t self, tb_char_t const* cstr, tb_size_t size, tb_size_t start, tb_char_t const* replace_cstr, tb_size_t replace_size, tb_size_t* plength)
{
// check
tb_regex_t* regex = (tb_regex_t*)self;
tb_assert_and_check_return_val(regex && regex->code && cstr && replace_cstr, tb_null);
// done
tb_char_t const* result = tb_null;
do
{
// clear length first
if (plength) *plength = 0;
// end?
tb_check_break(start < size);
// init options
#ifdef __tb_debug__
tb_uint32_t options = 0;
#else
tb_uint32_t options = PCRE2_NO_UTF_CHECK;
#endif
if (regex->mode & TB_REGEX_MODE_GLOBAL) options |= PCRE2_SUBSTITUTE_GLOBAL;
// init buffer
if (!regex->buffer_data)
{
regex->buffer_maxn = tb_max(size + replace_size + 64, 256);
regex->buffer_data = (PCRE2_UCHAR*)tb_malloc_bytes(regex->buffer_maxn);
}
tb_assert_and_check_break(regex->buffer_data);
// done
tb_long_t ok = -1;
PCRE2_SIZE length = 0;
while (1)
{
// replace it
length = (PCRE2_SIZE)regex->buffer_maxn;
ok = pcre2_substitute(regex->code, (PCRE2_SPTR)cstr, (PCRE2_SIZE)size, (PCRE2_SIZE)start, options, tb_null, tb_null, (PCRE2_SPTR)replace_cstr, (PCRE2_SIZE)replace_size, regex->buffer_data, &length);
// no space?
if (ok == PCRE2_ERROR_NOMEMORY)
{
// grow buffer
regex->buffer_maxn <<= 1;
regex->buffer_data = (PCRE2_UCHAR*)tb_ralloc_bytes(regex->buffer_data, regex->buffer_maxn);
tb_assert_and_check_break(regex->buffer_data);
}
// failed
else if (ok < 0)
{
#if defined(__tb_debug__) && !defined(TB_CONFIG_OS_WINDOWS)
// get error info
PCRE2_UCHAR info[256];
pcre2_get_error_message(ok, info, sizeof(info));
// trace
tb_trace_d("replace failed at offset %lu: error: %ld, %s\n", start, ok, info);
#endif
// end
break;
}
else break;
}
// check
tb_check_break(ok > 0);
tb_assert_and_check_break(length < regex->buffer_maxn);
// end
regex->buffer_data[length] = '\0';
// trace
tb_trace_d(" replace: [%lu]: %s", length, regex->buffer_data);
// save length
if (plength) *plength = (tb_size_t)length;
// ok
result = (tb_char_t const*)regex->buffer_data;
} while (0);
// ok?
return result;
}
tb_long_t tb_filter_spak(tb_filter_ref_t self, tb_byte_t const* data, tb_size_t size, tb_byte_t const** pdata, tb_size_t need, tb_long_t sync)
{
// check
tb_filter_t* filter = (tb_filter_t*)self;
tb_assert_and_check_return_val(filter && filter->spak && pdata, -1);
// init odata
*pdata = tb_null;
// save the input offset
filter->offset += size;
// eof?
if (filter->limit >= 0 && filter->offset == filter->limit)
filter->beof = tb_true;
// eof? sync it
if (filter->beof) sync = -1;
// the idata
tb_byte_t const* idata = tb_buffer_data(&filter->idata);
tb_size_t isize = tb_buffer_size(&filter->idata);
if (data && size)
{
// append data to cache if have the cache data
if (idata && isize)
{
// trace
tb_trace_d("[%p]: append idata: %lu", self, size);
// append data
idata = tb_buffer_memncat(&filter->idata, data, size);
isize = tb_buffer_size(&filter->idata);
}
// using the data directly if no cache data
else
{
// trace
tb_trace_d("[%p]: using idata directly: %lu", self, size);
// using it directly
idata = data;
isize = size;
}
}
// sync data if null
else
{
// check sync
tb_assert_and_check_return_val(sync, 0);
}
// the need
if (!need) need = tb_max(size, tb_queue_buffer_maxn(&filter->odata));
tb_assert_and_check_return_val(need, -1);
// init pull
tb_size_t omaxn = 0;
tb_byte_t* odata = tb_queue_buffer_pull_init(&filter->odata, &omaxn);
if (odata)
{
// the osize
tb_long_t osize = omaxn >= need? need : 0;
// exit pull
if (odata) tb_queue_buffer_pull_exit(&filter->odata, osize > 0? osize : 0);
// enough?
if (osize > 0)
{
// append to the cache if idata is not belong to the cache
if (size && idata == data) tb_buffer_memncat(&filter->idata, data, size);
// return it directly
*pdata = odata;
return osize;
}
}
// grow odata maxn if not enough
if (need > tb_queue_buffer_maxn(&filter->odata))
tb_queue_buffer_resize(&filter->odata, need);
// the odata
omaxn = 0;
odata = tb_queue_buffer_push_init(&filter->odata, &omaxn);
tb_assert_and_check_return_val(odata && omaxn, -1);
// init stream
tb_static_stream_t istream = {0};
tb_static_stream_t ostream = {0};
if (idata && isize)
{
// @note istream maybe null for sync the end data
if (!tb_static_stream_init(&istream, (tb_byte_t*)idata, isize)) return -1;
}
if (!tb_static_stream_init(&ostream, (tb_byte_t*)odata, omaxn)) return -1;
// trace
tb_trace_d("[%p]: spak: ileft: %lu, oleft: %lu, offset: %llu, limit: %lld, beof: %d: ..", self, tb_buffer_size(&filter->idata), tb_queue_buffer_size(&filter->odata), filter->offset, filter->limit, filter->beof);
// spak data
tb_long_t osize = filter->spak(filter, &istream, &ostream, sync);
// eof?
if (osize < 0) filter->beof = tb_true;
// no data and eof?
if (!osize && !tb_static_stream_left(&istream) && filter->beof) osize = -1;
// eof? sync it
if (filter->beof) sync = -1;
// exit odata
tb_queue_buffer_push_exit(&filter->odata, osize > 0? osize : 0);
// have the left idata?
tb_size_t left = tb_static_stream_left(&istream);
if (left)
{
// move to the cache head if idata is belong to the cache
if (idata != data)
{
// trace
tb_trace_d("[%p]: move to the cache head: %lu", self, left);
tb_buffer_memnmov(&filter->idata, tb_static_stream_offset(&istream), left);
}
// append to the cache if idata is not belong to the cache
else
{
// trace
tb_trace_d("[%p]: append to the cache: %lu", self, left);
tb_buffer_memncat(&filter->idata, tb_static_stream_pos(&istream), left);
}
}
// clear the cache
else tb_buffer_clear(&filter->idata);
// init pull
omaxn = 0;
odata = tb_queue_buffer_pull_init(&filter->odata, &omaxn);
// no sync? cache the output data
if (!sync) osize = omaxn >= need? need : 0;
// sync and has data? return it directly
else if (omaxn) osize = tb_min(omaxn, need);
// sync, no data or end?
// else osize = osize;
// exit pull
if (odata) tb_queue_buffer_pull_exit(&filter->odata, osize > 0? osize : 0);
// return it if have the odata
if (osize > 0) *pdata = odata;
// trace
tb_trace_d("[%p]: spak: ileft: %lu, oleft: %lu, offset: %llu, limit: %lld, beof: %d: %ld", self, tb_buffer_size(&filter->idata), tb_queue_buffer_size(&filter->odata), filter->offset, filter->limit, filter->beof, osize);
// ok?
return osize;
}
tb_void_t tb_pool_data_dump(tb_cpointer_t data, tb_bool_t verbose, tb_char_t const* prefix)
{
// done
tb_pool_data_head_t* data_head = tb_null;
do
{
// no data?
tb_assert_and_check_break(data);
// the data head
data_head = &(((tb_pool_data_head_t*)data)[-1]);
// dump the head info
tb_size_t data_limit = 256;
if (data_head->debug.magic == TB_POOL_DATA_MAGIC)
{
// the data size
tb_size_t data_size = (tb_size_t)data_head->size;
// format the backtrace prefix
tb_char_t backtrace_prefix[256] = {0};
tb_snprintf(backtrace_prefix, sizeof(backtrace_prefix), "%s ", prefix? prefix : "");
// dump backtrace
tb_size_t nframe = 0;
while (nframe < tb_arrayn(data_head->debug.backtrace) && data_head->debug.backtrace[nframe]) nframe++;
tb_trace_i("%sdata: from: %s(): %u, %s", prefix? prefix : "", data_head->debug.func, data_head->debug.line, data_head->debug.file);
tb_backtrace_dump(backtrace_prefix, data_head->debug.backtrace, nframe);
// dump the data info
tb_trace_i("%sdata: %p, size: %lu, patch: %x", prefix? prefix : "", data, data_size, ((tb_byte_t const*)data)[data_size]);
// dump the first 256-bytes data
if (data_size && verbose)
{
// the dump size
tb_size_t dump_size = tb_min(data_size, data_limit);
// dump it
tb_trace_i("%sdata: first %lu-bytes:", prefix? prefix : "", dump_size);
tb_pool_data_dump_data((tb_byte_t const*)data, dump_size);
// dump the last 256-bytes data
if (data_size > dump_size)
{
// the last data
tb_byte_t const* data_last = tb_max((tb_byte_t const*)data + data_size - data_limit, (tb_byte_t const*)data + dump_size);
// update the dump size
dump_size = (tb_byte_t const*)data + data_size - data_last;
// dump it
tb_trace_i("%sdata: last %lu-bytes:", prefix? prefix : "", dump_size);
tb_pool_data_dump_data(data_last, dump_size);
}
}
}
// for the public fixed_pool
else if (data_head->debug.magic == TB_POOL_DATA_EMPTY_MAGIC)
{
// format the backtrace prefix
tb_char_t backtrace_prefix[256] = {0};
tb_snprintf(backtrace_prefix, sizeof(backtrace_prefix), "%s ", prefix? prefix : "");
// dump backtrace
tb_size_t nframe = 0;
while (nframe < tb_arrayn(data_head->debug.backtrace) && data_head->debug.backtrace[nframe]) nframe++;
tb_trace_i("%sdata: from: %s(): %u, %s", prefix? prefix : "", data_head->debug.func, data_head->debug.line, data_head->debug.file);
tb_backtrace_dump(backtrace_prefix, data_head->debug.backtrace, nframe);
// dump the data info
tb_trace_i("%sdata: %p, size: fixed", prefix? prefix : "", data);
}
else
{
// dump the data head
tb_trace_i("%sdata: invalid head:", prefix? prefix : "");
tb_pool_data_dump_data((tb_byte_t const*)data_head, sizeof(tb_pool_data_head_t));
// dump the first 256-bytes data
tb_trace_i("%sdata: first %lu-bytes:", prefix? prefix : "", data_limit);
tb_pool_data_dump_data((tb_byte_t const*)data, data_limit);
}
} while (0);
}
int main (int argc, char *argv[])
{
int i, k = 0, n;
char *s;
char **optarg = NULL;
char *fn_in = NULL;
char *fn_out = NULL;
char *fn_app = NULL;
char *blanks = NULL;
char *fldseps = NULL;
char *recseps = NULL;
char *comment = NULL;
char *isep = " ";
char *impl = " <- ";
char *dflt = " (%1S)";
char *format = dflt;
int target = 's';
int min = 1;
int max = INT_MAX;
double supp = 0.1;
double smax = 1.0;
double conf = 0.8;
int dir = 0;
int eval = 0;
int aggm = 0;
double minval = 0.1;
int prune = 0;
double filter = 0.1;
int sort = 2;
int tree = 1;
int heap = 1;
int post = 0;
int report = 0;
int mode = APP_BODY|IST_PERFECT;
int size;
int wgt;
int frq, body, head;
int *items;
clock_t t, tt, tc, x;
#ifndef QUIET
prgname = argv[0];
if (argc > 1) {
fprintf(stderr, "%s - %s\n", argv[0], DESCRIPTION);
fprintf(stderr, VERSION); }
else {
printf("usage: %s [options] infile outfile\n", argv[0]);
printf("%s\n", DESCRIPTION);
printf("%s\n", VERSION);
printf("-t# target type "
"(default: %c)\n", target);
printf(" (s: frequent item sets, c: closed item sets,\n"
" m: maximal item sets, r: association rules)\n");
printf("-m# minimum number of items per set/rule "
"(default: %d)\n", min);
printf("-n# maximum number of items per set/rule "
"(default: no limit)\n");
printf("-s# minimum support of a set/rule "
"(default: %g%%)\n", supp *100);
printf("-S# maximum support of a set/rule "
"(default: %g%%)\n", smax *100);
printf(" (positive: percentage, "
"negative: absolute number)\n");
printf("-c# minimum confidence of a rule "
"(default: %g%%)\n", conf *100);
printf("infile file to read transactions from\n");
printf("outfile file to write item sets to\n");
return 0;
}
#endif
for (i = 1; i < argc; i++) {
s = argv[i];
if (optarg) { *optarg = s; optarg = NULL; continue; }
if ((*s == '-') && *++s) {
while (*s) {
switch (*s++) {
case '!': help(); break;
case 't': target = (*s) ? *s++ : 's'; break;
case 'm': min = (int)strtol(s, &s, 0); break;
case 'n': max = (int)strtol(s, &s, 0); break;
case 's': supp = 0.01*strtod(s, &s); break;
case 'S': smax = 0.01*strtod(s, &s); break;
case 'c': conf = 0.01*strtod(s, &s); break;
case 'o': mode |= APP_BOTH; break;
case 'e': eval = (*s) ? *s++ : 0; break;
case 'a': aggm = (*s) ? *s++ : 0; break;
case 'd': minval = 0.01*strtod(s, &s); break;
case 'p': prune = (int)strtol(s, &s, 0); break;
case 'g': report = ISR_SCAN; break;
case 'k': optarg = &isep; break;
case 'i': optarg = &impl; break;
case 'v': optarg = &format; break;
case 'l': dir = (int)strtol(s, &s, 0); break;
case 'q': sort = (int)strtol(s, &s, 0); break;
case 'u': filter = strtod(s, &s); break;
case 'h': tree = 0; break;
case 'j': heap = 0; break;
case 'x': mode &= ~IST_PERFECT; break;
case 'y': post = 1; break;
case 'b': optarg = &blanks; break;
case 'f': optarg = &fldseps; break;
case 'r': optarg = &recseps; break;
case 'C': optarg = &comment; break;
default : error(E_OPTION, *--s); break;
}
if (optarg && *s) { *optarg = s; optarg = NULL; break; }
} }
else {
switch (k++) {
case 0: fn_in = s; break;
case 1: fn_out = s; break;
case 2: fn_app = s; break;
default: error(E_ARGCNT); break;
}
}
}
if (optarg) error(E_OPTARG);
if ((k < 2) || (k > 3))
error(E_ARGCNT);
if ((!fn_in || !*fn_in) && (fn_app && !*fn_app))
error(E_STDIN);
switch (target) {
case 's': target = TT_SET; break;
case 'c': target = TT_CLOSED; break;
case 'm': target = TT_MAXIMAL; break;
case 'r': target = TT_RULE; break;
default : error(E_TARGET, (char)target); break;
}
if (min < 0) error(E_SIZE, min);
if (max < 0) error(E_SIZE, max);
if (supp > 1)
error(E_SUPP, supp);
if ((conf < 0) || (conf > 1))
error(E_CONF, conf);
switch (eval) {
case 'x': case 0: eval = IST_NONE; break;
case 'c': eval = IST_CONF; break;
case 'd': eval = IST_DIFF; break;
case 'l': eval = IST_LIFT; break;
case 'a': eval = IST_LD21; break;
case 'q': eval = IST_QUOT; break;
case 'n': eval = IST_CHI2; break;
case 'p': eval = IST_PVAL; break;
case 'i': eval = IST_INFO; break;
case 'g': eval = IST_PGST; break;
case 'b': eval = IST_LOGQ; break;
default : error(E_MEASURE, (char)eval); break;
}
switch (aggm) {
case 'x': case 0: aggm = IST_NONE; break;
case 'm': aggm = IST_MIN; break;
case 'n': aggm = IST_MAX; break;
case 'a': aggm = IST_AVG; break;
default : error(E_MEASURE, (char)aggm); break;
}
if ((target > TT_SET)
|| ((eval > IST_NONE) && (eval < IST_LOGQ)))
mode &= ~IST_PERFECT;
if (target <= TT_MAXIMAL) {
mode |= APP_BOTH; conf = 1;}
if ((filter <= -1) || (filter >= 1))
filter = 0;
ibase = ib_create(-1);
if (!ibase) error(E_NOMEM);
ib_chars(ibase, blanks, fldseps, recseps, comment);
MSG(stderr, "\n");
if (fn_app) {
t = clock();
if (*fn_app)
in = fopen(fn_app, "r");
else {
in = stdin; fn_app = "<stdin>"; }
MSG(stderr, "reading %s ... ", fn_app);
if (!in) error(E_FOPEN, fn_app);
k = ib_readapp(ibase, in);
if (k != 0) error(k, fn_app, RECCNT(ibase), BUFFER(ibase));
if (in != stdin) fclose(in);
in = NULL;
MSG(stderr, "[%d item(s)]", ib_cnt(ibase));
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
}
t = clock();
if (fn_in && *fn_in)
in = fopen(fn_in, "r");
else {
in = stdin; fn_in = "<stdin>"; }
MSG(stderr, "reading %s ... ", fn_in);
if (!in) error(E_FOPEN, fn_in);
tabag = tb_create(ibase);
if (!tabag) error(E_NOMEM);
while (1) {
k = ib_read(ibase, in);
if (k) { if (k > 0) break;
error(k, fn_in, RECCNT(ibase), BUFFER(ibase)); }
if (tb_add(tabag, NULL) != 0) error(E_NOMEM);
}
if (in != stdin) fclose(in);
in = NULL;
n = ib_cnt(ibase);
k = tb_cnt(tabag);
wgt = tb_wgt(tabag);
MSG(stderr, "[%d item(s), ", n);
if (k == wgt) MSG(stderr, "%d transaction(s)]", k);
else MSG(stderr, "%d/%d transaction(s)]", k, wgt);
MSG(stderr, " done [%.2fs].", SEC_SINCE(t));
if ((n <= 0) || (wgt <= 0))
error(E_NOTRANS);
MSG(stderr, "\n");
if (format == dflt) {
if (target != TT_RULE) format = (supp < 0) ? " (%a)" : " (%1S)";
else format = (supp < 0) ? " (%b, %1C)" : " (%1X, %1C)";
}
supp = ceil (((supp < 0) ? -100 : wgt) *supp);
smax = floor(((smax < 0) ? -100 : wgt) *smax);
t = clock();
MSG(stderr, "filtering, sorting and recoding items ... ");
map = (int*)malloc(n *sizeof(int));
if (!map) error(E_NOMEM);
k = (int)((mode & APP_HEAD) ? supp : ceil(supp *conf));
n = ib_recode(ibase, k, sort, map);
tb_recode(tabag, map);
tb_itsort(tabag, 1, heap);
free(map); map = NULL;
MSG(stderr, "[%d item(s)] done [%.2fs].", n, SEC_SINCE(t));
if (n <= 0) error(E_NOFREQ);
MSG(stderr, "\n");
k = tb_max(tabag);
if (max > k) max = k;
t = clock();
MSG(stderr, "reducing transactions ... ");
tb_filter(tabag, min, NULL);
tb_sort(tabag, 1, heap);
k = tb_reduce(tabag);
if (k == wgt) MSG(stderr, "[%d transaction(s)]", k);
else MSG(stderr, "[%d/%d transaction(s)]", k, wgt);
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
tt = 0;
if (tree) {
t = clock();
MSG(stderr, "building transaction tree ... ");
tatree = tt_create(tabag);
if (!tatree) error(E_NOMEM);
if (filter == 0) {
tb_delete(tabag, 0);
tabag = NULL;
}
MSG(stderr, "[%d node(s)]", tt_nodecnt(tatree));
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
tt = clock() -t;
}
t = clock(); tc = 0;
istree = ist_create(ibase, mode, (int)supp, (int)smax, conf);
if (!istree) error(E_NOMEM);
ist_seteval(istree, eval, aggm, minval, prune);
/* --- check item subsets --- */
MSG(stderr, "checking subsets of size 1");
map = (int*)malloc(n *sizeof(int));
if (!map) error(E_NOMEM);
while (1) {
size = ist_height(istree);
if (size >= max) break;
if ((filter != 0)
&& (ist_check(istree, map) <= size))
break;
if (post)
ist_prune(istree);
k = ist_addlvl(istree);
if (k) { if (k > 0) break;
error(E_NOMEM); }
if (((filter < 0)
&& (i < -filter *n))
|| ((filter > 0)
&& (i < n) && (i *(double)tt < filter *n *tc))) {
n = i;
x = clock();
tb_filter(tabag, size+1, map);
tb_sort(tabag, 0, heap);
tb_reduce(tabag);
if (tatree) {
tt_delete(tatree, 0);
tatree = tt_create(tabag);
if (!tatree) error(E_NOMEM);
}
tt = clock() -x;
}
MSG(stderr, " %d", ++size);
x = clock();
if (tatree) ist_countx(istree, tatree);
else ist_countb(istree, tabag);
tc = clock() -x;
}
free(map); map = NULL;
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
if ((target == TT_CLOSED) || (target == TT_MAXIMAL)) {
t = clock();
MSG(stderr, "filtering for %s item sets ... ",
(target == TT_MAXIMAL) ? "maximal" : "closed");
k = target | ((prune < 0) ? IST_EVAL : 0);
ist_mark(istree, k);
MSG(stderr, "done [%.2fs].\n", SEC_SINCE(t));
}
t = clock();
if (fn_out && *fn_out)
out = fopen(fn_out, "w");
else {
out = stdout; fn_out = "<stdout>"; }
MSG(stderr, "writing %s ... ", fn_out);
if (!out) error(E_FOPEN, fn_out);
if (eval == IST_LOGQ) report |= ISR_LOGS;
if ((target == TT_CLOSED) || (target == TT_MAXIMAL))
report |= ISR_CLOSED;
isrep = isr_create(ibase, out, report, isep, impl);
if (!isrep) error(E_NOMEM);
isr_setfmt (isrep, format);
isr_setsize(isrep, min, max);
ist_setsize(istree, min, max, dir);
ist_init (istree);
items = t_items(ib_tract(ibase));
if ((target <= TT_MAXIMAL)
&& (dir == 0)) {
if (eval == IST_LOGQ)
isr_seteval(isrep, isr_logq, NULL, minval);
else if (eval > IST_NONE)
isr_seteval(isrep, ist_evalx, istree, minval);
n = ist_report(istree, isrep); }
else if (target <= TT_MAXIMAL) {
for (n = 0; 1; ) {
k = ist_set(istree, items, &frq, &minval);
if (k < 0) break;
if (k > 0) fputs(isr_name(isrep, items[0]), out);
for (i = 0; ++i < k; ) {
fputs(isep, out); fputs(isr_name(isrep, items[i]), out); }
if (format)
isr_sinfo(isrep, frq, minval);
fputc('\n', out); n++;
} }
else if (target == TT_RULE) {
for (n = 0; 1; ) {
k = ist_rule(istree, items, &frq, &body, &head, &minval);
if (k < 0) break;
fputs(isr_name(isrep, items[0]), out);
fputs(impl, out);
if (k > 1) fputs(isr_name(isrep, items[1]), out);
for (i = 1; ++i < k; ) {
fputs(isep, out); fputs(isr_name(isrep, items[i]), out); }
if (format)
isr_rinfo(isrep, frq, body, head, minval);
fputc('\n', out); n++;
}
} /
if (fflush(out) != 0) error(E_FWRITE, fn_out);
int main (int argc, char *argv[])
{ /* --- main function */
int i, k = 0, n; /* loop variables, counters */
char *s; /* to traverse the options */
char **optarg = NULL; /* option argument */
char *fn_in = NULL; /* name of input file */
char *fn_out = NULL; /* name of output file */
char *fn_app = NULL; /* name of item appearances file */
char *blanks = NULL; /* blanks */
char *fldseps = NULL; /* field separators */
char *recseps = NULL; /* record separators */
char *comment = NULL; /* comment indicators */
char *isep = " "; /* item separator for output */
char *impl = " <- "; /* implication sign for ass. rules */
char *dflt = " (%1S)"; /* default format for check */
char *format = dflt; /* format for information output */
int target = 's'; /* target type (sets/rules/h.edges) */
int min = 1; /* minimum rule/item set size */
int max = INT_MAX; /* maximum rule/item set size */
double supp = 10; /* minimum support (in percent) */
double smax = 100; /* maximum support (in percent) */
double conf = 80; /* minimum confidence (in percent) */
int dir = 0; /* direction for size sorting */
int eval = 0; /* additional evaluation measure */
int zero = 0; /* flag for zero eval. below expect. */
int aggm = 0; /* aggregation mode for eval. measure */
double minval = 10; /* minimum evaluation measure value */
int prune = 0; /* (min. size for) evaluation pruning */
double filter = 0.1; /* item usage filtering parameter */
int sort = 2; /* flag for item sorting and recoding */
int tree = 1; /* flag for transaction tree */
int heap = 1; /* flag for heap sort vs. quick sort */
int post = 0; /* flag for a-posteriori pruning */
int report = 0; /* other flags for reporting */
int mode = APP_BODY|IST_PERFECT; /* search mode */
int size; /* current item set size */
int wgt; /* total transaction weight */
int frq, body, head; /* frequency of an item set */
int *items; /* item set (for reporting) */
clock_t t, tt, tc, x; /* timers for measurements */
#ifndef QUIET /* if not quiet version */
prgname = argv[0]; /* get program name for error msgs. */
/* --- print usage message --- */
if (argc > 1) { /* if arguments are given */
fprintf(stderr, "%s - %s\n", argv[0], DESCRIPTION);
fprintf(stderr, VERSION); } /* print a startup message */
else { /* if no arguments are given */
printf("usage: %s [options] infile outfile [appfile]\n", argv[0]);
printf("%s\n", DESCRIPTION);
printf("%s\n", VERSION);
printf("-t# target type "
"(default: %c)\n", target);
printf(" (s: frequent item sets, c: closed item sets,\n"
" m: maximal item sets, r: association rules)\n");
printf("-m# minimum number of items per set/rule "
"(default: %d)\n", min);
printf("-n# maximum number of items per set/rule "
"(default: no limit)\n");
printf("-s# minimum support of a set/rule "
"(default: %g%%)\n", supp);
printf("-S# maximum support of a set/rule "
"(default: %g%%)\n", smax);
printf(" (positive: percentage, "
"negative: absolute number)\n");
printf("-c# minimum confidence of a rule "
"(default: %g%%)\n", conf);
printf("-o use the original rule support definition "
"(body & head)\n");
printf("-e# additional evaluation measure "
"(default: none)\n");
printf("-a# aggregation mode for evaluation measure "
"(default: none)\n");
printf("-z zero evaluation below expected support "
"(default: evaluate all)\n");
printf("-d# minimum value of add. evaluation measure "
"(default: %g%%)\n", minval);
printf("-p# (min. size for) pruning with evaluation "
"(default: no pruning)\n");
printf(" (< 0: backward, > 0: forward)\n");
printf("-l# sort item sets in output by their size "
"(default: no sorting)\n");
printf(" (< 0: descending, > 0: ascending)\n");
printf("-g write item names in scanable form "
"(quote certain characters)\n");
printf("-k# item separator for output "
"(default: \"%s\")\n", isep);
printf("-i# implication sign for association rules "
"(default: \"%s\")\n", impl);
printf("-v# output format for set/rule information "
"(default: \"%s\")\n", format);
printf("-q# sort items w.r.t. their frequency "
"(default: %d)\n", sort);
printf(" (1: ascending, -1: descending, 0: do not sort,\n"
" 2: ascending, -2: descending w.r.t. "
"transaction size sum)\n");
printf("-u# filter unused items from transactions "
"(default: %g)\n", filter);
printf(" (0: do not filter items w.r.t. usage in sets,\n"
" <0: fraction of removed items for filtering,\n"
" >0: take execution times ratio into account)\n");
printf("-j use quicksort to sort the transactions "
"(default: heapsort)\n");
printf("-x do not prune the search "
"with perfect extensions\n");
printf("-y a-posteriori pruning of infrequent item sets\n");
printf("-h do not organize transactions as a prefix tree\n");
printf("-b# blank characters "
"(default: \" \\t\\r\")\n");
printf("-f# field separators "
"(default: \" \\t,\")\n");
printf("-r# record separators "
"(default: \"\\n\")\n");
printf("-C# comment characters "
"(default: \"#\")\n");
printf("-! print additional option information\n");
printf("infile file to read transactions from\n");
printf("outfile file to write item sets/association rules"
"/hyperedges to\n");
printf("appfile file stating item appearances (optional)\n");
return 0; /* print a usage message */
} /* and abort the program */
#endif /* #ifndef QUIET */
/* free option characters: w [A-Z]\[SC] */
/* --- evaluate arguments --- */
for (i = 1; i < argc; i++) { /* traverse the arguments */
s = argv[i]; /* get an option argument */
if (optarg) { *optarg = s; optarg = NULL; continue; }
if ((*s == '-') && *++s) { /* -- if argument is an option */
while (*s) { /* traverse the options */
switch (*s++) { /* evaluate the options */
case '!': help(); break;
case 't': target = (*s) ? *s++ : 's'; break;
case 'm': min = (int)strtol(s, &s, 0); break;
case 'n': max = (int)strtol(s, &s, 0); break;
case 's': supp = strtod(s, &s); break;
case 'S': smax = strtod(s, &s); break;
case 'c': conf = strtod(s, &s); break;
case 'o': mode |= APP_BOTH; break;
case 'e': eval = (*s) ? *s++ : 0; break;
case 'z': zero = IST_ZERO; break;
case 'a': aggm = (*s) ? *s++ : 0; break;
case 'd': minval = strtod(s, &s); break;
case 'p': prune = (int)strtol(s, &s, 0); break;
case 'g': report = ISR_SCAN; break;
case 'k': optarg = &isep; break;
case 'i': optarg = &impl; break;
case 'v': optarg = &format; break;
case 'l': dir = (int)strtol(s, &s, 0); break;
case 'q': sort = (int)strtol(s, &s, 0); break;
case 'u': filter = strtod(s, &s); break;
case 'h': tree = 0; break;
case 'j': heap = 0; break;
case 'x': mode &= ~IST_PERFECT; break;
case 'y': post = 1; break;
case 'b': optarg = &blanks; break;
case 'f': optarg = &fldseps; break;
case 'r': optarg = &recseps; break;
case 'C': optarg = &comment; break;
default : error(E_OPTION, *--s); break;
} /* set the option variables */
if (optarg && *s) { *optarg = s; optarg = NULL; break; }
} } /* get an option argument */
else { /* -- if argument is no option */
switch (k++) { /* evaluate non-options */
case 0: fn_in = s; break;
case 1: fn_out = s; break;
case 2: fn_app = s; break;
default: error(E_ARGCNT); break;
} /* note filenames */
}
}
if (optarg) error(E_OPTARG); /* check the option argument */
if ((k < 2) || (k > 3)) /* and the number of arguments */
error(E_ARGCNT); /* (either in/out or in/out/app) */
if ((!fn_in || !*fn_in) && (fn_app && !*fn_app))
error(E_STDIN); /* stdin must not be used twice */
switch (target) { /* check and translate target type */
case 's': target = TT_ALL; break;
case 'c': target = TT_CLOSED; break;
case 'm': target = TT_MAXIMAL; break;
case 'r': target = TT_RULE; break;
default : error(E_TARGET, (char)target); break;
}
if (min < 0) error(E_SIZE, min); /* check the limits */
if (max < 0) error(E_SIZE, max); /* for the set size */
if (supp > 100) /* check the minimum support */
error(E_SUPP, supp); /* (< 0: absolute support) */
if ((conf < 0) || (conf > 100))
error(E_CONF, conf); /* check the minimum confidence */
switch (eval) { /* check and translate measure */
case 'x': case 0: eval = IST_NONE; break;
case 'c': eval = IST_CONF; break;
case 'd': eval = IST_CONF_DIFF; break;
case 'l': eval = IST_LIFT; break;
case 'a': eval = IST_LIFT_DIFF; break;
case 'q': eval = IST_LIFT_QUOT; break;
case 'v': eval = IST_CVCT; break;
case 'e': eval = IST_CVCT_DIFF; break;
case 'r': eval = IST_CVCT_QUOT; break;
case 'f': eval = IST_CERT; break;
case 'n': eval = IST_CHI2; break;
case 'p': eval = IST_CHI2_PVAL; break;
case 'i': eval = IST_INFO; break;
case 'g': eval = IST_INFO_PVAL; break;
case 'b': eval = IST_LOGQ; break;
default : error(E_MEASURE, (char)eval); break;
}
switch (aggm) { /* check and translate agg. mode */
case 'x': case 0: aggm = IST_NONE; break;
case 'm': aggm = IST_MIN; break;
case 'n': aggm = IST_MAX; break;
case 'a': aggm = IST_AVG; break;
default : error(E_MEASURE, (char)aggm); break;
}
if ((target > TT_ALL) /* if individual set counters needed */
|| ((eval > IST_NONE) && (eval < IST_LOGQ)))
mode &= ~IST_PERFECT; /* remove perfect extension pruning */
if (target <= TT_MAXIMAL) { /* remove rule specific settings */
mode |= APP_BOTH; conf = 100; }
if ((filter <= -1) || (filter >= 1))
filter = 0; /* check and adapt the filter option */
if (dir) /* if to sort output by size, */
mode &= ~IST_PERFECT; /* do not use perfect ext. pruning */
/* --- create item base --- */
ibase = ib_create(0, 0); /* create an item base and */
if (!ibase) error(E_NOMEM); /* set the special characters */
ib_chars(ibase, blanks, fldseps, recseps, "", comment);
MSG(stderr, "\n"); /* terminate the startup message */
/* --- read item appearance indicators --- */
if (fn_app) { /* if item appearances are given */
t = clock(); /* start the timer for the reading */
if (*fn_app) /* if an app. file name is given, */
in = fopen(fn_app, "r"); /* open the item appearances file */
else { /* if no app. file name is given, */
in = stdin; fn_app = "<stdin>"; } /* read from std. input */
MSG(stderr, "reading %s ... ", fn_app);
if (!in) error(E_FOPEN, fn_app);
k = ib_readapp(ibase, in); /* read the item appearances */
if (k != 0) error(k, fn_app, RECCNT(ibase), BUFFER(ibase));
if (in != stdin) fclose(in);/* if not read from standard input, */
in = NULL; /* close the input file */
MSG(stderr, "[%d item(s)]", ib_cnt(ibase));
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
} /* print a log message */
/* --- read transactions --- */
t = clock(); /* start the timer for the reading */
if (fn_in && *fn_in) /* if an input file name is given, */
in = fopen(fn_in, "r"); /* open input file for reading */
else { /* if no input file name is given, */
in = stdin; fn_in = "<stdin>"; } /* read from standard input */
MSG(stderr, "reading %s ... ", fn_in);
if (!in) error(E_FOPEN, fn_in);
tabag = tb_create(ibase, 0); /* create a transaction bag/multiset */
if (!tabag) error(E_NOMEM); /* to store the transactions */
while (1) { /* transaction read loop */
k = ib_read(ibase, in); /* read the next transaction */
if (k) { if (k > 0) break; /* check for error and end of file */
error(k, fn_in, RECCNT(ibase), BUFFER(ibase)); }
if (tb_add(tabag, NULL) != 0) error(E_NOMEM);
} /* add transaction to bag/multiset */
if (in != stdin) fclose(in); /* if not read from standard input, */
in = NULL; /* close the input file */
n = ib_cnt(ibase); /* get the number of items */
k = tb_cnt(tabag); /* get the number of transactions */
wgt = tb_wgt(tabag); /* the total transaction weight */
MSG(stderr, "[%d item(s), ", n);
if (k == wgt) MSG(stderr, "%d transaction(s)]", k);
else MSG(stderr, "%d/%d transaction(s)]", k, wgt);
MSG(stderr, " done [%.2fs].", SEC_SINCE(t));
if ((n <= 0) || (wgt <= 0)) /* check for at least one item */
error(E_NOTRANS); /* and at least one transaction */
MSG(stderr, "\n"); /* terminate the log message */
if (format == dflt) { /* if default info. format is used */
if (target != TT_RULE) format = (supp < 0) ? " (%a)" : " (%1S)";
else format = (supp < 0) ? " (%b, %1C)" : " (%1X, %1C)";
} /* set default according to target */
supp = ceil ((supp >= 0) ? 0.01 *supp *wgt : -supp);
smax = floor((smax >= 0) ? 0.01 *smax *wgt : -smax);
conf *= 0.01; /* transform support and confidence */
/* --- sort and recode items --- */
t = clock(); /* compute absolute support values */
MSG(stderr, "filtering, sorting and recoding items ... ");
map = (int*)malloc(n *sizeof(int));
if (!map) error(E_NOMEM); /* create an item identifier map */
k = (int)((mode & APP_HEAD) ? supp : ceil(supp *conf));
n = ib_recode(ibase, k, sort, map);
tb_recode(tabag, map); /* recode the items and transactions */
tb_itsort(tabag, 1, heap); /* and sort items in transactions */
free(map); map = NULL; /* delete the item identifier map */
MSG(stderr, "[%d item(s)] done [%.2fs].", n, SEC_SINCE(t));
if (n <= 0) error(E_NOFREQ); /* print a log message and */
MSG(stderr, "\n"); /* check the number of items */
k = tb_max(tabag); /* clamp the set/rule length to */
if (max > k) max = k; /* the maximum transaction size */
/* --- reduce transactions --- */
t = clock(); /* start the timer for the reduction */
MSG(stderr, "reducing transactions ... ");
tb_filter(tabag, min, NULL); /* remove items of short transactions */
tb_sort(tabag, 1, heap); /* sort the trans. lexicographically */
k = tb_reduce(tabag); /* reduce transactions to unique ones */
if (k == wgt) MSG(stderr, "[%d transaction(s)]", k);
else MSG(stderr, "[%d/%d transaction(s)]", k, wgt);
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
/* --- create transaction tree --- */
tt = 0; /* init. the tree construction time */
if (tree) { /* if to use a transaction tree */
t = clock(); /* start the timer for construction */
MSG(stderr, "building transaction tree ... ");
tatree = tt_create(tabag); /* create a transaction tree */
if (!tatree) error(E_NOMEM);
if (filter == 0) { /* if not to filter items, */
tb_delete(tabag, 0); /* delete the transaction bag */
tabag = NULL; /* (redundant data storage) */
}
MSG(stderr, "[%d node(s)]", tt_nodecnt(tatree));
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
tt = clock() -t; /* note the time for the construction */
} /* of the transaction tree */
/* --- create item set tree --- */
t = clock(); tc = 0; /* start the timer for the search */
istree = ist_create(ibase, mode, (int)supp, (int)smax, conf);
if (!istree) error(E_NOMEM); /* create an item set tree */
ist_seteval(istree, eval|zero, aggm, 0.01*minval, prune);
/* --- check item subsets --- */
MSG(stderr, "checking subsets of size 1");
map = (int*)malloc(n *sizeof(int));
if (!map) error(E_NOMEM); /* create a filter map */
while (1) { /* traverse the item set sizes */
size = ist_height(istree); /* get the current item set size and */
if (size >= max) break; /* abort if maximal size is reached */
if ((filter != 0) /* if to filter w.r.t. item usage */
&& (ist_check(istree, map) <= size))
break; /* check which items are still used */
if (post) /* if a-posteriori pruning requested, */
ist_prune(istree); /* prune infrequent item sets */
k = ist_addlvl(istree); /* while max. height is not reached, */
if (k) { if (k > 0) break; /* add a level to the item set tree */
error(E_NOMEM); } /* if no level was added, abort */
if (((filter < 0) /* if to filter w.r.t. item usage */
&& (i < -filter *n)) /* and enough items were removed */
|| ((filter > 0) /* or counting time is long enough */
&& (i < n) && (i *(double)tt < filter *n *tc))) {
n = i; /* note the new number of items */
x = clock(); /* start the timer for filtering */
tb_filter(tabag, size+1, map);
tb_sort(tabag, 0, heap); /* remove unnec. items and trans. */
tb_reduce(tabag); /* and reduce trans. to unique ones */
if (tatree) { /* if a transaction tree was created */
tt_delete(tatree, 0); /* delete the transaction tree */
tatree = tt_create(tabag);
if (!tatree) error(E_NOMEM);
} /* rebuild the transaction tree */
tt = clock() -x; /* note the filter/rebuild time */
}
MSG(stderr, " %d", ++size); /* print the current item set size */
x = clock(); /* start the timer for counting */
if (tatree) ist_countx(istree, tatree);
else ist_countb(istree, tabag);
tc = clock() -x; /* count the transaction tree/bag */
} /* and compute the new counting time */
free(map); map = NULL; /* delete the filter map */
MSG(stderr, " done [%.2fs].\n", SEC_SINCE(t));
/* --- filter found item sets --- */
if ((target == TT_CLOSED) || (target == TT_MAXIMAL)) {
t = clock(); /* start the timer for filtering */
MSG(stderr, "filtering for %s item sets ... ",
(target == TT_MAXIMAL) ? "maximal" : "closed");
k = target | ((prune < 0) ? IST_EVAL : 0);
ist_mark(istree, k); /* filter closed/maximal item sets */
MSG(stderr, "done [%.2fs].\n", SEC_SINCE(t));
} /* print a log message */
/* --- print item sets/rules/hyperedges --- */
t = clock(); /* start the timer for the output */
if (fn_out && *fn_out) /* if an output file name is given, */
out = fopen(fn_out, "w"); /* open the output file */
else { /* if no output file name is given, */
out = stdout; fn_out = "<stdout>"; } /* write to std. output */
MSG(stderr, "writing %s ... ", fn_out);
if (!out) error(E_FOPEN, fn_out);
if (eval == IST_LOGQ) report |= ISR_LOGS;
if ((target == TT_CLOSED) || (target == TT_MAXIMAL))
report |= ISR_NOEXP; /* combine the report mode flags */
isrep = isr_create(ibase, out, report, isep, impl);
if (!isrep) error(E_NOMEM); /* create an item set reporter */
isr_setfmt (isrep, format); /* and configure it */
isr_setsize(isrep, min, max);
ist_setsize(istree, min, max, dir);
ist_init (istree); /* initialize the extraction */
items = t_items(ib_tract(ibase));
if ((target <= TT_MAXIMAL) /* if to find frequent item sets */
&& (dir == 0)) { /* and not to sort them by size */
if (eval == IST_LOGQ) /* if to compute an add. evaluation */
isr_seteval(isrep, isr_logq, NULL, 0.01*minval);
else if (eval > IST_NONE) /* set the add. evaluation function */
isr_seteval(isrep, ist_evalx, istree, 0.01*minval);
n = ist_report(istree, isrep); } /* report the item sets */
else if (target <= TT_MAXIMAL) { /* if to find frequent item sets */
for (n = 0; 1; ) { /* extract item sets from the tree */
k = ist_set(istree, items, &frq, &minval);
if (k < 0) break; /* get the next frequent item set */
if (k > 0) fputs(isr_name(isrep, items[0]), out);
for (i = 0; ++i < k; ) { /* print the item names */
fputs(isep, out); fputs(isr_name(isrep, items[i]), out); }
if (format) /* if requested, print information */
isr_sinfo(isrep, frq, minval);
fputc('\n', out); n++; /* terminate the output line and */
} } /* count the reported item set */
else if (target == TT_RULE) { /* if to find association rules, */
for (n = 0; 1; ) { /* extract rules from tree */
k = ist_rule(istree, items, &frq, &body, &head, &minval);
if (k < 0) break; /* get the next association rule */
fputs(isr_name(isrep, items[0]), out);
fputs(impl, out); /* print name of rule head item */
if (k > 1) fputs(isr_name(isrep, items[1]), out);
for (i = 1; ++i < k; ) { /* print names of items in rule body */
fputs(isep, out); fputs(isr_name(isrep, items[i]), out); }
if (format) /* if requested, print information */
isr_rinfo(isrep, frq, body, head, minval);
fputc('\n', out); n++; /* terminate the output line and */
} /* count the reported ass. rule */
} /* if (target <= TT_MAXIMAL) .. else .. */
if (fflush(out) != 0) error(E_FWRITE, fn_out);
if (out != stdout) fclose(out);
out = NULL; /* close the output file */
MSG(stderr, "[%d %s(s)] done ", n,
(target == TT_RULE) ? "rule" : "set");
MSG(stderr, "[%.2fs].\n", SEC_SINCE(t));
#ifdef BENCH /* if benchmark version, */
ist_stats(istree); /* show the search statistics */
#endif /* (especially memory usage) */
/* --- clean up --- */
#ifndef NDEBUG /* if this is a debug version */
isr_delete(isrep, 0); /* the item set reporter, */
ist_delete(istree); /* the item set tree, */
if (tatree) tt_delete(tatree, 0); /* the transaction tree, */
if (tabag) tb_delete(tabag, 0); /* the transaction bag, */
ib_delete(ibase); /* and the item base */
#endif
#ifdef STORAGE /* if storage debugging */
showmem("at end of program"); /* check memory usage */
#endif
return 0; /* return 'ok' */
} /* main() */
