diff --git a/.editorconfig b/.editorconfig index e23a3c7..e9c1e63 100644 --- a/.editorconfig +++ b/.editorconfig @@ -5,6 +5,5 @@ root = true end_of_line = lf insert_final_newline = true -[*.v] -# vfmt wants it :( +[*.{v,c,h}] indent_style = tab diff --git a/src/cron/expression/c/expression.c b/src/cron/expression/c/expression.c index f9dc534..f15e359 100644 --- a/src/cron/expression/c/expression.c +++ b/src/cron/expression/c/expression.c @@ -4,15 +4,15 @@ const uint8_t month_days[] = {31, 28, 31, 30, 31, 30, 31, 31, 30, 31, 30, 31}; struct cron_expression *ce_init() { - return malloc(sizeof(struct cron_expression)); + return malloc(sizeof(struct cron_expression)); } void ce_free(struct cron_expression *ce) { - free(ce->months); - free(ce->days); - free(ce->hours); - free(ce->minutes); - free(ce); + free(ce->months); + free(ce->days); + free(ce->hours); + free(ce->minutes); + free(ce); } int ce_next(struct cron_simple_time *out, struct cron_expression *ce, struct cron_simple_time *ref) { @@ -21,100 +21,101 @@ int ce_next(struct cron_simple_time *out, struct cron_expression *ce, struct cro // means we've looped back around. This means that the "bigger" value has // to be incremented by one. For example, if the minutes have looped // around, that means that the hour has to be incremented as well. - uint8_t month_index = 0; - uint8_t day_index = 0; - uint8_t hour_index = 0; - uint8_t minute_index = 0; + uint8_t month_index = 0; + uint8_t day_index = 0; + uint8_t hour_index = 0; + uint8_t minute_index = 0; // This chain is the same logic multiple times, namely that if a "bigger" // value loops around, then the smaller value will always reset as well. // For example, if we're going to a new day, the hour & minute will always // be their smallest value again. - while (month_index < ce->month_count && ref->month > ce->months[month_index]) { - month_index++; - } + while (month_index < ce->month_count && ref->month > ce->months[month_index]) { + month_index++; + } - if (month_index < ce->month_count && ref->month == ce->months[month_index]) { - while (day_index < ce->day_count && ref->day > ce->days[day_index]) { - day_index++; - } + if (month_index < ce->month_count && ref->month == ce->months[month_index]) { + while (day_index < ce->day_count && ref->day > ce->days[day_index]) { + day_index++; + } - if (day_index < ce->day_count && ref->day == ce->days[day_index]) { - while (hour_index < ce->hour_count && ref->hour > ce->hours[hour_index]) { - hour_index++; - } + if (day_index < ce->day_count && ref->day == ce->days[day_index]) { + while (hour_index < ce->hour_count && ref->hour > ce->hours[hour_index]) { + hour_index++; + } - if (hour_index < ce->hour_count && ref->hour == ce->hours[hour_index]) { + if (hour_index < ce->hour_count && ref->hour == ce->hours[hour_index]) { // Minute is the only value where we explicitely make sure we // can't match sref's value exactly. This is to ensure we only // return values in the future. - while (minute_index < ce->minute_count && ref->minute >= ce->minutes[minute_index]) { - minute_index++; - } - } - } - } + while (minute_index < ce->minute_count && ref->minute >= ce->minutes[minute_index]) { + minute_index++; + } + } + } + } // Here, we increment the "bigger" values by one if the smaller ones loop // around. The order is important, as it allows a sort-of waterfall effect // to occur which updates all values if required. - if (minute_index == ce->minute_count && hour_index < ce->hour_count) { - hour_index++; - } + if (minute_index == ce->minute_count && hour_index < ce->hour_count) { + hour_index++; + } - if (hour_index == ce->hour_count && day_index < ce->day_count) { - day_index++; - } + if (hour_index == ce->hour_count && day_index < ce->day_count) { + day_index++; + } - if (day_index == ce->day_count && month_index < ce->month_count) { - month_index++; - } + if (day_index == ce->day_count && month_index < ce->month_count) { + month_index++; + } - out->minute = ce->minutes[minute_index % ce->minute_count]; - out->hour = ce->hours[hour_index % ce->hour_count]; - out->day = ce->days[day_index % ce->day_count]; + out->minute = ce->minutes[minute_index % ce->minute_count]; + out->hour = ce->hours[hour_index % ce->hour_count]; + out->day = ce->days[day_index % ce->day_count]; // Sometimes, we end up with a day that does not exist within the selected // month, e.g. day 30 in February. When this occurs, we reset day back to // the smallest value & loop over to the next month that does have this // day. - if (out->day > month_days[ce->months[month_index % ce->month_count] - 1]) { - out->day = ce->days[0]; - month_index++; + if (out->day > month_days[ce->months[month_index % ce->month_count] - 1]) { + out->day = ce->days[0]; + month_index++; - while (out->day > month_days[ce->months[month_index % ce->month_count] - 1]) { - month_index++; - - if (month_index == 2 * ce->month_count) { - return 1; - } - } - } + while (out->day > month_days[ce->months[month_index % ce->month_count] - 1]) { + month_index++; - out->month = ce->months[month_index % ce->month_count]; + // TODO find out if this can happen + if (month_index == 2 * ce->month_count) { + return 1; + } + } + } - if (month_index >= ce->month_count) { - out->year = ref->year + 1; - } else { - out->year = ref->year; - } + out->month = ce->months[month_index % ce->month_count]; - return 0; + if (month_index >= ce->month_count) { + out->year = ref->year + 1; + } else { + out->year = ref->year; + } + + return 0; } int ce_next_from_now(struct cron_simple_time *out, struct cron_expression *ce) { - time_t t = time(NULL); - struct tm gm; - gmtime_r(&t, &gm); + time_t t = time(NULL); + struct tm gm; + gmtime_r(&t, &gm); - struct cron_simple_time ref = { - .year = gm.tm_year, - // tm_mon goes from 0 to 11 - .month = gm.tm_mon + 1, - .day = gm.tm_mday, - .hour = gm.tm_hour, - .minute = gm.tm_min - }; + struct cron_simple_time ref = { + .year = gm.tm_year, + // tm_mon goes from 0 to 11 + .month = gm.tm_mon + 1, + .day = gm.tm_mday, + .hour = gm.tm_hour, + .minute = gm.tm_min + }; - return ce_next(out, ce, &ref); + return ce_next(out, ce, &ref); } diff --git a/src/cron/expression/c/expression.h b/src/cron/expression/c/expression.h index 91db5d0..f6d8826 100644 --- a/src/cron/expression/c/expression.h +++ b/src/cron/expression/c/expression.h @@ -4,29 +4,29 @@ #include enum cron_parse_error { - CPEParseOk = 0, - CPEParseInvalidExpression = 1, - CPEParseInvalidNumber = 2, - CPEParseOutOfRange = 3 + CPEParseOk = 0, + CPEParseInvalidExpression = 1, + CPEParseInvalidNumber = 2, + CPEParseOutOfRange = 3 }; struct cron_expression { - uint8_t *minutes; - uint8_t *hours; - uint8_t *days; - uint8_t *months; - uint8_t minute_count; - uint8_t hour_count; - uint8_t day_count; - uint8_t month_count; + uint8_t *minutes; + uint8_t *hours; + uint8_t *days; + uint8_t *months; + uint8_t minute_count; + uint8_t hour_count; + uint8_t day_count; + uint8_t month_count; }; struct cron_simple_time { - int year; - int month; - int day; - int hour; - int minute; + int year; + int month; + int day; + int hour; + int minute; }; struct cron_expression *ce_init(); diff --git a/src/cron/expression/c/parse.c b/src/cron/expression/c/parse.c index e664dd8..cd23458 100644 --- a/src/cron/expression/c/parse.c +++ b/src/cron/expression/c/parse.c @@ -1,25 +1,28 @@ #include "expression.h" +// Allowed value ranges for the minute, hour, day and month field const uint8_t min[4] = {0, 0, 1, 1}; const uint8_t max[4] = {59, 23, 31, 12}; +// Convert a string a uint8_t value by parsing it using atoi and checking +// whether it's contained within the given range #define SAFE_ATOI(v,s,min,max) \ - int _##v = atoi(s); \ - if ((_##v) == 0 && strcmp((s), "0") != 0) { \ - return CPEParseInvalidNumber; \ - } \ - if (v < (min) || v > (max)) { \ - return CPEParseOutOfRange; \ - } \ - v = (uint8_t) (_##v); + int _##v = atoi(s); \ + if ((_##v) == 0 && strcmp((s), "0") != 0) { \ + return CPEParseInvalidNumber; \ + } \ + if (v < (min) || v > (max)) { \ + return CPEParseOutOfRange; \ + } \ + v = (uint8_t) (_##v); /** * Given a range expression, produce a bit field defining what numbers in the - * min-max range the expression represents. The first bit (starting from the - * right) corresponds to min, the max - min + 1'th bit to max. All trailing bits + * min-max range the expression represents. Bit 0 (starting from the + * right) corresponds to min, the bit max - min to max. All trailing bits * after this should be ignored. The given bitfield is modified in-place, so * multiple calls of this function can be performed on the same value to create - * the effect of ORing their values: + * the effect of ORing their values. * * A range expression has one of the following forms: * @@ -30,185 +33,210 @@ const uint8_t max[4] = {59, 23, 31, 12}; * - a-b/c */ enum cron_parse_error ce_parse_range(uint64_t *out, char *s, uint8_t min, uint8_t max) { - // The * expression means "every possible value" - if (s[0] == '*') { - // A '*' is only valid on its own - if (s[1] != '\0') { - return CPEParseInvalidExpression; - } + // The * expression means "every possible value" + if (s[0] == '*') { + // A '*' is only valid on its own + if (s[1] != '\0') { + return CPEParseInvalidExpression; + } - *out = ~0; + *out = ~0; - return CPEParseOk; - } + return CPEParseOk; + } - size_t slash_index = 0; - size_t dash_index = 0; - size_t i = 0; + size_t slash_index = 0; + size_t dash_index = 0; + size_t i = 0; - // We first iterate over the string to determine whether it contains a slash - // and/or a dash. We know the dash can only be valid if it appears before - // the slash. - while (s[i] != '\0' && slash_index == 0) { - if (s[i] == '/') { - slash_index = i; + // We first iterate over the string to determine whether it contains a slash + // and/or a dash. We know the dash can only be valid if it appears before + // the slash. + while (s[i] != '\0' && slash_index == 0) { + if (s[i] == '/') { + slash_index = i; - s[i] = '\0'; - } else if (s[i] == '-') { - dash_index = i; + s[i] = '\0'; + } else if (s[i] == '-') { + dash_index = i; - s[i] = '\0'; - } + s[i] = '\0'; + } - i++; - } + i++; + } - // Parse the three possible numbers in the pattern - uint8_t start = 0; - uint8_t end = 0; - uint8_t interval = 1; + // Parse the three possible numbers in the pattern + uint8_t start = 0; + uint8_t end = max; + uint8_t interval = 1; - SAFE_ATOI(start, s, min, max); + SAFE_ATOI(start, s, min, max); - if (dash_index > 0) { - SAFE_ATOI(end, &s[dash_index + 1], min, max); - } + if (dash_index > 0) { + SAFE_ATOI(end, &s[dash_index + 1], min, max); + } - if (slash_index > 0) { - SAFE_ATOI(interval, &s[slash_index + 1], 1, max - min); - } + if (slash_index > 0) { + SAFE_ATOI(interval, &s[slash_index + 1], 1, max - min); + } - // Single number doesn't need to loop - if (end == 0 && slash_index == 0) { - *out |= ((uint64_t) 1) << (start - min); - } else { - for (;start <= end; start += interval) { - *out |= ((uint64_t) 1) << (start - min); - start += interval; - } - } + if (dash_index == 0 && slash_index == 0) { + *out |= ((uint64_t) 1) << (start - min); + } else { + while (start <= end) { + *out |= ((uint64_t) 1) << (start - min); + start += interval; + } + } - return CPEParseOk; + return CPEParseOk; } +/* + * Given an expression part, produce a bitfield defining what numbers in the + * min-max range the part represents. A part consists of one or more range + * expressions, separated by commas. + */ enum cron_parse_error ce_parse_part(uint64_t *out, char *s, uint8_t min, uint8_t max) { - *out = 0; + *out = 0; - char *next; - enum cron_parse_error res; - - while ((next = strchr(s, ',')) != NULL) { - next[0] = '\0'; - res = ce_parse_range(out, s, min, max); + char *next; + enum cron_parse_error res; - if (res != CPEParseOk) { - return res; - } + while ((next = strchr(s, ',')) != NULL) { + next[0] = '\0'; + res = ce_parse_range(out, s, min, max); - s = next + 1; - } + if (res != CPEParseOk) { + return res; + } - // Make sure to parse the final range as well - return ce_parse_range(out, s, min, max); + s = next + 1; + } + + // Make sure to parse the final range as well + return ce_parse_range(out, s, min, max); } +/* + * Return how many bits are set in the bitfield, better known as popcount. I + * added my own implementation (taken from my algorithms course) as I don't want + * to be dependent on GCC-specific extensions. + */ +uint8_t uint64_t_popcount(uint64_t n) { + uint8_t c = 0; + + while (n != 0) { + // This sets the least significant bit to zero (very cool) + n &= n - 1; + c++; + } + + return c; +} + +/* + * Convert a bitfield into an array containing the numbers in the min-max range + * it represents. + */ uint8_t bf_to_nums(uint8_t **out, uint64_t bf, uint8_t min, uint8_t max) { - uint8_t capacity = 8; - uint8_t size = 0; + // Each bit field only has `max - min + 1` meaningful bits. All other bits + // should be ignored, and can be any value. By shifting the bit field back and + // forth, we set these excessive bits to zero, ensuring popcount returns the + // correct value. + uint8_t excess_bits = 64 - (max - min + 1); + bf = (bf << excess_bits) >> excess_bits; + uint8_t size = uint64_t_popcount(bf); + uint8_t *buf = malloc(size * sizeof(uint8_t)); - uint8_t *buf = malloc(capacity * sizeof(uint8_t)); + uint8_t i = 0, j = 0; - for (uint8_t i = 0; i <= max - min; i++) { - if (((uint64_t) 1 << i) & bf) { - // Resize buffer if needed - if (size == capacity) { - capacity *= 2; - buf = realloc(buf, capacity * sizeof(uint8_t)); - } + while (j < size && i <= max - min) { + if (((uint64_t)1 << i) & bf) { + // Resize buffer if needed + buf[j] = min + i; + j++; + } - buf[size] = min + i; - size++; - } - } + i++; + } - // Resize buffer once more to remove any trailing unused bytes - if (size < capacity) { - buf = realloc(buf, size * sizeof(uint8_t)); - } + *out = buf; - *out = buf; - - return size; + return size; } +/* + * Parse a cron expression string into a cron_expression struct. + */ enum cron_parse_error ce_parse_expression(struct cron_expression *out, char *s) { - // The parsing functions modify the input string in-place - s = strdup(s); - char *orig_s = s; + // The parsing functions modify the input string in-place + s = strdup(s); + char *orig_s = s; - uint8_t part_count = 0; + uint8_t part_count = 0; - char *next; - enum cron_parse_error res = CPEParseOk; - uint64_t bfs[4]; + char *next; + enum cron_parse_error res = CPEParseOk; + uint64_t bfs[4]; - // Skip leading spaces - while (s[0] == ' ') { - s++; - } - - while (part_count < 4 && ((next = strchr(s, ' ')) != NULL)) { - next[0] = '\0'; - res = ce_parse_part(&bfs[part_count], s, min[part_count], max[part_count]); + // Skip leading spaces + while (s[0] == ' ') { + s++; + } - if (res != CPEParseOk) { - goto end; - } + while (part_count < 4 && ((next = strchr(s, ' ')) != NULL)) { + next[0] = '\0'; + res = ce_parse_part(&bfs[part_count], s, min[part_count], max[part_count]); - size_t offset = 1; + if (res != CPEParseOk) { + goto end; + } - // Skip multiple spaces - while (next[offset] == ' ') { - offset++; - } - s = next + offset; + size_t offset = 1; - part_count++; - } + // Skip multiple spaces + while (next[offset] == ' ') { + offset++; + } + s = next + offset; - // Parse final trailing part - if (part_count < 4 && s[0] != '\0') { - // Make sure to parse the final range as well - res = ce_parse_part(&bfs[part_count], s, min[part_count], max[part_count]); + part_count++; + } - if (res != CPEParseOk) { - goto end; - } + // Parse final trailing part + if (part_count < 4 && s[0] != '\0') { + res = ce_parse_part(&bfs[part_count], s, min[part_count], max[part_count]); - part_count++; - } + if (res != CPEParseOk) { + goto end; + } - // At least two parts need to be provided - if (part_count < 2) { - res = CPEParseInvalidExpression; - goto end; - } + part_count++; + } - // Ensure there's always 4 parts, as expressions can have between 2 and 4 parts - while (part_count < 4) { - // Expression is augmented with '*' expressions - bfs[part_count] = ~0; - part_count++; - } + // At least two parts need to be provided + if (part_count < 2) { + res = CPEParseInvalidExpression; + goto end; + } - out->minute_count = bf_to_nums(&out->minutes, bfs[0], min[0], max[0]); - out->hour_count = bf_to_nums(&out->hours, bfs[1], min[1], max[1]); - out->day_count = bf_to_nums(&out->days, bfs[2], min[2], max[2]); - out->month_count = bf_to_nums(&out->months, bfs[3], min[3], max[3]); + // Ensure there's always 4 parts, as expressions can have between 2 and 4 parts + while (part_count < 4) { + // Expression is augmented with '*' expressions + bfs[part_count] = ~0; + part_count++; + } + + out->minute_count = bf_to_nums(&out->minutes, bfs[0], min[0], max[0]); + out->hour_count = bf_to_nums(&out->hours, bfs[1], min[1], max[1]); + out->day_count = bf_to_nums(&out->days, bfs[2], min[2], max[2]); + out->month_count = bf_to_nums(&out->months, bfs[3], min[3], max[3]); end: - // s is cloned - free(orig_s); + // s is cloned + free(orig_s); - return res; + return res; } diff --git a/src/cron/expression/expression_test.v b/src/cron/expression/expression_test.v index 2b21b4b..448927a 100644 --- a/src/cron/expression/expression_test.v +++ b/src/cron/expression/expression_test.v @@ -22,15 +22,15 @@ fn test_next_simple() ! { /* util_test_time('0 3', '2002-01-01 00:00:00', '2002-01-01 03:00:00')! */ // Overlap to next day - mut exp := '0 3' + mut exp := '0 3 ' util_test_time(exp, '2002-01-01 03:00:00', '2002-01-02 03:00:00')! util_test_time(exp, '2002-01-01 04:00:00', '2002-01-02 03:00:00')! - /* util_test_time('0 3/4', '2002-01-01 04:00:00', '2002-01-01 07:00:00')! */ + util_test_time('0 3/4', '2002-01-01 04:00:00', '2002-01-01 07:00:00')! /* // Overlap to next month */ - /* util_test_time('0 3', '2002-11-31 04:00:00', '2002-12-01 03:00:00')! */ + util_test_time('0 3', '2002-11-31 04:00:00', '2002-12-01 03:00:00')! /* // Overlap to next year */ - /* util_test_time('0 3', '2002-12-31 04:00:00', '2003-01-01 03:00:00')! */ + util_test_time('0 3', '2002-12-31 04:00:00', '2003-01-01 03:00:00')! }