// Match.new needle, string, options = {}
VALUE CommandTMatch_initialize(int argc, VALUE *argv, VALUE self)
{
// process arguments: 2 mandatory, 1 optional
VALUE str, needle, options;
if (rb_scan_args(argc, argv, "21", &str, &needle, &options) == 2)
options = Qnil;
str = StringValue(str);
needle = StringValue(needle); // already downcased by caller
// check optional options hash for overrides
VALUE always_show_dot_files = CommandT_option_from_hash("always_show_dot_files", options);
VALUE never_show_dot_files = CommandT_option_from_hash("never_show_dot_files", options);
matchinfo_t m;
m.haystack_p = RSTRING_PTR(str);
m.haystack_len = RSTRING_LEN(str);
m.needle_p = RSTRING_PTR(needle);
m.needle_len = RSTRING_LEN(needle);
m.max_score_per_char = (1.0 / m.haystack_len + 1.0 / m.needle_len) / 2;
m.dot_file = 0;
m.always_show_dot_files = always_show_dot_files == Qtrue;
m.never_show_dot_files = never_show_dot_files == Qtrue;
// calculate score
double score = 1.0;
// special case for zero-length search string
if (m.needle_len == 0) {
// filter out dot files
if (!m.always_show_dot_files) {
for (long i = 0; i < m.haystack_len; i++) {
char c = m.haystack_p[i];
if (c == '.' && (i == 0 || m.haystack_p[i - 1] == '/')) {
score = 0.0;
break;
}
}
}
} else if (m.haystack_len > 0) { // normal case
// prepare for memoization
double memo[m.haystack_len * m.needle_len];
for (long i = 0, max = m.haystack_len * m.needle_len; i < max; i++)
memo[i] = DBL_MAX;
m.memo = memo;
score = recursive_match(&m, 0, 0, 0, 0.0);
}
// clean-up and final book-keeping
rb_iv_set(self, "@score", rb_float_new(score));
rb_iv_set(self, "@str", str);
return Qnil;
}
void calculate_match(VALUE str,
VALUE needle,
VALUE case_sensitive,
VALUE always_show_dot_files,
VALUE never_show_dot_files,
VALUE recurse,
match_t *out)
{
long i, max;
double score;
matchinfo_t m;
m.haystack_p = RSTRING_PTR(str);
m.haystack_len = RSTRING_LEN(str);
m.needle_p = RSTRING_PTR(needle);
m.needle_len = RSTRING_LEN(needle);
m.max_score_per_char = (1.0 / m.haystack_len + 1.0 / m.needle_len) / 2;
m.always_show_dot_files = always_show_dot_files == Qtrue;
m.never_show_dot_files = never_show_dot_files == Qtrue;
m.case_sensitive = case_sensitive;
m.recurse = recurse == Qtrue;
// calculate score
score = 1.0;
// special case for zero-length search string
if (m.needle_len == 0) {
// filter out dot files
if (!m.always_show_dot_files) {
for (i = 0; i < m.haystack_len; i++) {
char c = m.haystack_p[i];
if (c == '.' && (i == 0 || m.haystack_p[i - 1] == '/')) {
score = 0.0;
break;
}
}
}
} else if (m.haystack_len > 0) { // normal case
// prepare for memoization
double memo[m.haystack_len * m.needle_len];
for (i = 0, max = m.haystack_len * m.needle_len; i < max; i++)
memo[i] = DBL_MAX;
m.memo = memo;
score = recursive_match(&m, 0, 0, 0, 0.0);
}
// final book-keeping
out->path = str;
out->score = score;
}
double recursive_match(matchinfo_t *m, // sharable meta-data
long haystack_idx, // where in the path string to start
long needle_idx, // where in the needle string to start
long last_idx, // location of last matched character
double score) // cumulative score so far
{
double score_for_char;
double seen_score = 0; // remember best score seen via recursion
int found;
long i, j, distance;
long memo_idx = haystack_idx;
// do we have a memoized result we can return?
double memoized = m->memo[needle_idx * m->needle_len + memo_idx];
if (memoized != DBL_MAX)
return memoized;
// bail early if not enough room (left) in haystack for (rest of) needle
if (m->haystack_len - haystack_idx < m->needle_len - needle_idx) {
score = 0.0;
goto memoize;
}
for (i = needle_idx; i < m->needle_len; i++) {
char c = m->needle_p[i];
found = 0;
// similar to above, we'll stop iterating when we know we're too close
// to the end of the string to possibly match
for (j = haystack_idx;
j <= m->haystack_len - (m->needle_len - i);
j++, haystack_idx++) {
char d = m->haystack_p[j];
if (d == '.') {
if (j == 0 || m->haystack_p[j - 1] == '/') { // this is a dot-file
int dot_search = (i == 0 && c == '.'); // searching for a dot
if (m->never_show_dot_files || (!dot_search && !m->always_show_dot_files)) {
score = 0.0;
goto memoize;
}
}
} else if (d >= 'A' && d <= 'Z' && !m->case_sensitive) {
d += 'a' - 'A'; // add 32 to downcase
}
if (c == d) {
found = 1;
// calculate score
score_for_char = m->max_score_per_char;
distance = j - last_idx;
if (distance > 1) {
double factor = 1.0;
char last = m->haystack_p[j - 1];
char curr = m->haystack_p[j]; // case matters, so get again
if (last == '/')
factor = 0.9;
else if (last == '-' ||
last == '_' ||
last == ' ' ||
(last >= '0' && last <= '9'))
factor = 0.8;
else if (last >= 'a' && last <= 'z' &&
curr >= 'A' && curr <= 'Z')
factor = 0.8;
else if (last == '.')
factor = 0.7;
else
// if no "special" chars behind char, factor diminishes
// as distance from last matched char increases
factor = (1.0 / distance) * 0.75;
score_for_char *= factor;
}
if (++j < m->haystack_len) {
// bump cursor one char to the right and
// use recursion to try and find a better match
double sub_score = recursive_match(m, j, i, last_idx, score);
if (sub_score > seen_score)
seen_score = sub_score;
}
score += score_for_char;
last_idx = haystack_idx++;
break;
}
}
if (!found) {
score = 0.0;
goto memoize;
}
}
score = score > seen_score ? score : seen_score;
memoize:
m->memo[needle_idx * m->needle_len + memo_idx] = score;
return score;
}
void match_stdin(char *abbrev) {
char *line = 0;
size_t read = 0;
size_t max_line_len = 0;
size_t current_line_len = 0;
size_t results_buf_len = 10000;
matchresult_t *results_buf = malloc(results_buf_len * sizeof(matchresult_t));
size_t results_count = 0;
while ((read = getline(&line, ¤t_line_len, stdin)) != -1) {
matchinfo_t matchinfo = {
line,
strlen(line),
abbrev,
strlen(abbrev),
1.0, 0, 1, 0
};
double score = 0;
char *line_copy = 0;
// chomp newline at the end of filename
line[read - 1] = 0;
// if getline call expanded line buffer, remember new buffer size
if (current_line_len > max_line_len)
max_line_len = current_line_len;
// next iteration of while will reuse line buffer
current_line_len = max_line_len;
// do match
score = recursive_match(&matchinfo, 0, 0, 0, 0.0);
// record result only if score is higher than 0
if (score > 0.0) {
// realloc results_buf if needed
if (results_buf_len == results_count) {
results_buf = realloc(results_buf, (results_buf_len *= 2) * sizeof(matchresult_t));
}
// make copy of line
line_copy = malloc(read + 1);
memcpy(line_copy, line, read + 1);
// add result to results_buf
results_buf[results_count].line = line_copy;
results_buf[results_count].score = score;
results_count++;
}
}
// Sorting results
qsort(results_buf, results_count, sizeof(matchresult_t), matchresult_comp_func);
// Print sorted results
for (size_t i = 0; i < results_count; i++)
printf("%f: %s\n", results_buf[i].score, results_buf[i].line);
// Cleanup
free(line);
for (size_t i = 0; i < results_count; i++)
free(results_buf[i].line);
free(results_buf);
}
float recursive_match(
matchinfo_t *m, // Sharable meta-data.
long haystack_idx, // Where in the path string to start.
long needle_idx, // Where in the needle string to start.
long last_idx, // Location of last matched character.
float score // Cumulative score so far.
) {
long distance, i, j;
float *memoized = NULL;
float score_for_char;
float seen_score = 0;
// Iterate over needle.
for (i = needle_idx; i < m->needle_len; i++) {
// Iterate over (valid range of) haystack.
for (j = haystack_idx; j <= m->rightmost_match_p[i]; j++) {
char c, d;
// Do we have a memoized result we can return?
memoized = &m->memo[j * m->needle_len + i];
if (*memoized != UNSET_SCORE) {
return *memoized > seen_score ? *memoized : seen_score;
}
c = m->needle_p[i];
d = m->haystack_p[j];
if (d == '.') {
if (j == 0 || m->haystack_p[j - 1] == '/') { // This is a dot-file.
int dot_search = c == '.'; // Searching for a dot.
if (
m->never_show_dot_files ||
(!dot_search && !m->always_show_dot_files)
) {
return *memoized = 0.0;
}
}
} else if (d >= 'A' && d <= 'Z' && !m->case_sensitive) {
d += 'a' - 'A'; // Add 32 to downcase.
}
if (c == d) {
// Calculate score.
float sub_score = 0;
score_for_char = m->max_score_per_char;
distance = j - last_idx;
if (distance > 1) {
float factor = 1.0;
char last = m->haystack_p[j - 1];
char curr = m->haystack_p[j]; // Case matters, so get again.
if (last == '/') {
factor = 0.9;
} else if (
last == '-' ||
last == '_' ||
last == ' ' ||
(last >= '0' && last <= '9')
) {
factor = 0.8;
} else if (
last >= 'a' && last <= 'z' &&
curr >= 'A' && curr <= 'Z'
) {
factor = 0.8;
} else if (last == '.') {
factor = 0.7;
} else {
// If no "special" chars behind char, factor diminishes
// as distance from last matched char increases.
factor = (1.0 / distance) * 0.75;
}
score_for_char *= factor;
}
if (j < m->rightmost_match_p[i] && m->recurse) {
sub_score = recursive_match(m, j + 1, i, last_idx, score);
if (sub_score > seen_score) {
seen_score = sub_score;
}
}
last_idx = j;
haystack_idx = last_idx + 1;
score += score_for_char;
*memoized = seen_score > score ? seen_score : score;
if (i == m->needle_len - 1) {
// Whole string matched.
return *memoized;
}
}
}
}
return *memoized = score;
}
float calculate_match(
VALUE haystack,
VALUE needle,
VALUE case_sensitive,
VALUE always_show_dot_files,
VALUE never_show_dot_files,
VALUE recurse,
long needle_bitmask,
long *haystack_bitmask
) {
matchinfo_t m;
long i;
float score = 1.0;
int compute_bitmasks = *haystack_bitmask == UNSET_BITMASK;
m.haystack_p = RSTRING_PTR(haystack);
m.haystack_len = RSTRING_LEN(haystack);
m.needle_p = RSTRING_PTR(needle);
m.needle_len = RSTRING_LEN(needle);
m.rightmost_match_p = NULL;
m.max_score_per_char = (1.0 / m.haystack_len + 1.0 / m.needle_len) / 2;
m.always_show_dot_files = always_show_dot_files == Qtrue;
m.never_show_dot_files = never_show_dot_files == Qtrue;
m.case_sensitive = (int)case_sensitive;
m.recurse = recurse == Qtrue;
// Special case for zero-length search string.
if (m.needle_len == 0) {
// Filter out dot files.
if (m.never_show_dot_files || !m.always_show_dot_files) {
for (i = 0; i < m.haystack_len; i++) {
char c = m.haystack_p[i];
if (c == '.' && (i == 0 || m.haystack_p[i - 1] == '/')) {
return 0.0;
}
}
}
} else {
long haystack_limit;
long memo_size;
long needle_idx;
long mask;
long rightmost_match_p[m.needle_len];
if (*haystack_bitmask != UNSET_BITMASK) {
if ((needle_bitmask & *haystack_bitmask) != needle_bitmask) {
return 0.0;
}
}
// Pre-scan string:
// - Bail if it can't match at all.
// - Record rightmost match for each character (prune search space).
// - Record bitmask for haystack to speed up future searches.
m.rightmost_match_p = rightmost_match_p;
needle_idx = m.needle_len - 1;
mask = 0;
for (i = m.haystack_len - 1; i >= 0; i--) {
char c = m.haystack_p[i];
char lower = c >= 'A' && c <= 'Z' ? c + ('a' - 'A') : c;
if (!m.case_sensitive) {
c = lower;
}
if (compute_bitmasks) {
mask |= (1 << (lower - 'a'));
}
if (needle_idx >= 0) {
char d = m.needle_p[needle_idx];
if (c == d) {
rightmost_match_p[needle_idx] = i;
needle_idx--;
}
}
}
if (compute_bitmasks) {
*haystack_bitmask = mask;
}
if (needle_idx != -1) {
return 0.0;
}
// Prepare for memoization.
haystack_limit = rightmost_match_p[m.needle_len - 1] + 1;
memo_size = m.needle_len * haystack_limit;
{
float memo[memo_size];
for (i = 0; i < memo_size; i++) {
memo[i] = UNSET_SCORE;
}
m.memo = memo;
score = recursive_match(&m, 0, 0, 0, 0.0);
}
}
return score;
}
