/* Copyright (c) 1994 by Sanjay Ghemawat */

#include <ctype.h>
#include <string.h>

#include "../types/Array.h"

#include "Date.h"
#include "Month.h"
#include "WeekDay.h"
#include "Year.h"
#include "config.h"

#ifndef HAVE_STRNCASECMP_PROTO
extern "C" {
    extern int strncasecmp(char const*, char const*, unsigned long);
}
#endif

// Token types
enum Token_Type {
    INT_TOKEN,
    MONTH_TOKEN,
    WEEKDAY_TOKEN,

    TODAY_TOKEN,
    TOMORROW_TOKEN,
    YESTERDAY_TOKEN,

    AMPM_TOKEN,
    MIDNIGHT_TOKEN,
    NOON_TOKEN,
    ZONE_TOKEN,

    SLASH_TOKEN,
    COLON_TOKEN,

    TO_TOKEN,

    OTHER_TOKEN,
    LAST_TOKEN
}; 

// Part of string being parsed.
struct Token {
    char const*	ptr;		// Pointer into string
    int		len;		// Length of this token
    Token_Type	ttype;		// The type of this token
    long	tvalue;		// Any value associated with this token
};

declareArray(Tokens,Token)
implementArray(Tokens,Token)

static void parse_tokens(char const*, Tokens&);
static void parse_token(char const*, Token&);
static int  parse_word(char const*, int, Token&);

static void parse_tokens(char const* str, Tokens& list) {
    while (str[0] != '\0') {
	Token t;
	parse_token(str, t);
	str += t.len;

	// Merge trailing whitespace etc. into token
	while (str[0] != '\0') {
	    char c = str[0];
	    if ((c != '.') && (c != ',') && !isspace(c))
		break;

	    t.len++;
	    str++;
	}

	list.append(t);
    }

    // Append guard token
    Token t;
    t.ptr = str;
    t.len = 0;
    t.ttype = LAST_TOKEN;
    list.append(t);
}

// effects	Parse token from string and stuff it in "t".
static void parse_token(char const* string, Token& t) {
    char c = string[0];

    // Parse number
    if (isdigit(c)) {
	char* endptr;
	long v = strtol(string, &endptr, 10);
	if (endptr > string) {
	    t.ptr = string;
	    t.len = endptr - string;
	    t.ttype = INT_TOKEN;
	    t.tvalue = v;
	    
	    return;
	}
    }

    // Special handling for "am/pm".
    if (isalpha(c)) {
	// Skip periods and alphabetic characters.
	int end = 1;
	while ((string[end] != '\0') &&
	       (isalpha(string[end]) || (string[end] == '.')))
	    end++;

	if (parse_word(string, end, t))
	    return;
    }
    
    // Parse word
    if (isalpha(c)) {
	int end = 1;
	while ((string[end] != '\0') && isalpha(string[end]))
	    end++;

	if (parse_word(string, end, t))
	    return;

	// If word has exactly three uppercase letters, consider it a
	// time zone.
	if ((end == 3) &&
	    isupper(string[0]) &&
	    isupper(string[1]) &&
	    isupper(string[2])) {
	    t.ptr = string;
	    t.len = end;
	    t.ttype = ZONE_TOKEN;
	    return;
	}
    }

    // Parse zone
    if ((c == '+') || (c == '-')) {
	Token rest;
	parse_token(string+1, rest);
	if ((rest.ttype == INT_TOKEN) && (rest.len >= 3)) {
	    // Found a time zone
	    t.ptr = string;
	    t.len = rest.len+1;
	    t.ttype = ZONE_TOKEN;
	    return;
	}
    }

    // Parse range separator
    for (int seplen = 1; seplen <= 3; seplen++) {
	if (strncasecmp(string, "---", seplen) == 0) {
	    t.ptr = string;
	    t.len = seplen;
	    t.ttype = TO_TOKEN;
	    return;
	}
    }

    t.ptr = string;
    t.len = 1;

    switch (c) {
      case '/':
	t.ttype = SLASH_TOKEN;
	break;
      case ':':
	t.ttype = COLON_TOKEN;
	break;
      default:
	t.ttype = OTHER_TOKEN;
	break;
    }
}

// The lengths of the following tokens are the minimum lengths
// that are required for a match.
static Token words[] = {
    { "today",		5,	TODAY_TOKEN,	0	},
    { "tomorrow",	5,	TOMORROW_TOKEN,	0	},
    { "yesterday",	5,	YESTERDAY_TOKEN,0	},

    { "january",	3,	MONTH_TOKEN,	1	},
    { "february",	3,	MONTH_TOKEN,	2	},
    { "march",		3,	MONTH_TOKEN,	3	},
    { "april",		3,	MONTH_TOKEN,	4	},
    { "may",		3,	MONTH_TOKEN,	5	},
    { "june",		3,	MONTH_TOKEN,	6	},
    { "july",		3,	MONTH_TOKEN,	7	},
    { "august",		3,	MONTH_TOKEN,	8	},
    { "september",	3,	MONTH_TOKEN,	9	},
    { "october",	3,	MONTH_TOKEN,	10	},
    { "november",	3,	MONTH_TOKEN,	11	},
    { "december",	3,	MONTH_TOKEN,	12	},

    { "sunday",		3,	WEEKDAY_TOKEN,	1	},
    { "monday",		3,	WEEKDAY_TOKEN,	2	},
    { "tuesday",	3,	WEEKDAY_TOKEN,	3	},
    { "wednesday",	3,	WEEKDAY_TOKEN,	4	},
    { "thursday",	3,	WEEKDAY_TOKEN,	5	},
    { "friday",		3,	WEEKDAY_TOKEN,	6	},
    { "saturday",	3,	WEEKDAY_TOKEN,	7	},

    { "a.m.",		4,	AMPM_TOKEN,	0	},
    { "am.",		3,	AMPM_TOKEN,	0	},
    { "am",		2,	AMPM_TOKEN,	0	},
    { "p.m.",		4,	AMPM_TOKEN,	1	},
    { "pm.",		3,	AMPM_TOKEN,	1	},
    { "pm",		2,	AMPM_TOKEN,	1	},

    { "midnight",	8,	MIDNIGHT_TOKEN,	0	},
    { "noon",		4,	NOON_TOKEN,	0	},

    { "to",		2,	TO_TOKEN,	0	},

    { 0,		0,	OTHER_TOKEN,	0	}
};

static int build_date(int m, int d, int y, Date& result) {
    // Fix year
    if (y < 100) y += (Date::Today().GetYear() / 100) * 100;

    // Check range
    if ((y < Year::First()) || (y > Year::Last())) return 0;
    if ((m < 1) || (m > 12)) return 0;

    Month month = Month::First() + (m - 1);
    if ((d < 1) || (d > month.Size(y))) return 0;

    result = Date(d, month, y);
    return 1;
}

static int parse_word(char const* str, int len, Token& t) {
    for (int i = 0; words[i].ptr != 0; i++) {
	if (len < words[i].len) continue;
	if (len > strlen(words[i].ptr)) continue;
	if (strncasecmp(str, words[i].ptr, len) != 0) continue;

	// Found it
	t.ptr = str;
	t.len = len;
	t.ttype = words[i].ttype;

	t.tvalue = words[i].tvalue;
	return 1;
    }

    return 0;
}

// Used to ignore "am/pm" indicator from "match_timeofday"
static int junk_ampm;

// modifies	"result", "had_ampm"
// effects	If "list" starting at "index" contains a time specification,
//		parse the time into "result" and return the number of
//		tokens used in the parse.  Otherwise return zero.
//
//		"result" is set to the number of seconds after midnight.
//		"had_ampm" is set to true if an am/pm indication was
//		found, otherwise it is set to false.

static int match_timeofday(Tokens& list, int index, int& result,
			   int& had_ampm = junk_ampm) {
    if (index >= list.size()) return 0;
    Token t = list[index];

    if (t.ttype == MIDNIGHT_TOKEN) {
	result = 0;
	had_ampm = 1;
	return 1;
    }

    if (t.ttype == NOON_TOKEN) {
	result = 12 * 60 * 60;
	had_ampm = 1;
	return 1;
    }

    // "<hour>:<minute>[:<second>] [am|pm] [<time zone>]"
    if ((list[index+0].ttype == INT_TOKEN) &&
	(list[index+1].ttype == COLON_TOKEN) &&
	(list[index+2].ttype == INT_TOKEN)) {
	int h = list[index+0].tvalue;
	int m = list[index+2].tvalue;
	int s = 0;
	int c = 3;

	// See if seconds are specified
	if ((list[index+c+0].ttype == COLON_TOKEN) &&
	    (list[index+c+1].ttype == INT_TOKEN)) {
	    s = list[index+c+1].tvalue;
	    c += 2;
	}

	// See if am/pm is specified
	int have_ampm = 0;
	int ispm = 0;
	if (list[index+c].ttype == AMPM_TOKEN) {
	    have_ampm = 1;
	    ispm = list[index+c].tvalue;
	    c++;
	}

	// See if time zone is specified
	if (list[index+c].ttype == ZONE_TOKEN) {
	    // XXX - Handle zone
	    c++;
	}

	if (have_ampm) {
	    // Hours > 12 not allowed with am/pm
	    if (h > 12) return 0;

	    if (h == 12) h = 0;
	    if (ispm) h += 12;
	}

	// Range checking
	if (h >= 24) return 0;
	if ((m < 0) || (m > 60)) return 0;
	if ((s < 0) || (s > 60)) return 0;

	result = ((h * 60) + m) * 60 + s;
	had_ampm = (have_ampm || (h >= 12));
	return c;
    }

    // "<hour> [am|pm]"
    if ((list[index+0].ttype == INT_TOKEN) &&
	(list[index+1].ttype == AMPM_TOKEN)) {
	int h = list[index].tvalue;
	if (h > 12) return 0;
	if (h == 12) h = 0;
	if (list[index+1].tvalue) h += 12;

	result = (h * 60 * 60);
	had_ampm = 1;
	return 2;
    }

    return 0;
}

// effects	If "list" starting at "index" contains a range specification,
//		parse the range into and return the number of
//		tokens used in the parse.  Otherwise return zero.

static int match_timerange(Tokens& list, int index, int& start, int& finish) {
    if (index >= list.size()) return 0;
    Token t = list[index];
    int noon = 12 * 60 * 60;

    // "<timeofday> <to> <timeofday>"
    int c1, r1, c2, r2, ampm1, ampm2;
    if (((c1 = match_timeofday(list, index, r1, ampm1)) != 0) &&
	(list[index+c1].ttype == TO_TOKEN) &&
	((c2 = match_timeofday(list, index+c1+1, r2, ampm2)) != 0)) {

	if (!ampm1 && !ampm2 && (r1 > r2) && (r1 < noon) && (r2 < noon)) {
	    // r1 must be am and r2 must be pm
	    r2 += noon;
	}
	if (ampm1 && !ampm2 && (r1 < noon) && (r2 < r1)) {
	    // r2 must be pm
	    r2 += noon;
	}
	if (!ampm1 && ampm2 && (r1 < noon) && (r2 >= noon) &&
	    ((r1 + noon) <= r2)) {
	    // Make r1 match r2
	    r1 += noon;
	}

	if (r1 <= r2) {
	    start = r1;
	    finish = r2;
	    return c1 + c2 + 1;
	}
    }

    // "<hour> <to> <timeofday>"
    if ((list[index+0].ttype == INT_TOKEN) &&
	(list[index+1].ttype == TO_TOKEN) &&
	(list[index+0].tvalue > 0) &&
	(list[index+0].tvalue <= 12) &&
	((c2 = match_timeofday(list, index+2, r2, ampm2)) != 0)) {
	int h = list[index+0].tvalue;
	if (h == 12) h = 0;
	r1 = h * 60 * 60;

	if (!ampm2 && (r1 > r2) && (r2 < noon)) {
	    // r2 must be after noon
	    r2 += noon;
	}

	// Move "r1" into second half of day if range > 12 hrs
	if ((r2 - r1) > (12*60*60))
	    r1 += 12*60*60;

	if (r1 <= r2) {
	    start = r1;
	    finish = r2;
	    return c2 + 2;
	}
    }

    return 0;
}

// effects	If "list" starting at "index" contains a date specification,
//		parse the date into "date" and return the number of tokens used
//		in the parse.  Otherwise return zero.

static int match_date(Tokens& list, int index, Date& result) {
    if (index >= list.size()) return 0;

    Token t = list[index];

    // "<weekday> <date>"
    if (t.ttype == WEEKDAY_TOKEN) {
	// Try to find date starting at next token
	Date r;
	int count = match_date(list, index+1, r);
	if ((count > 0) && (r.GetWDay().Index() == t.tvalue)) {
	    // Weekday matches the date
	    result = r;
	    return count + 1;
	}

	// Weekday does not match
	return 0;
    }

    // "today"
    if (t.ttype == TODAY_TOKEN) {
	result = Date::Today();
	return 1;
    }

    // "tomorrow"
    if (t.ttype == TOMORROW_TOKEN) {
	result = Date::Today() + 1;
	return 1;
    }

    // "yesterday"
    if (t.ttype == YESTERDAY_TOKEN) {
	result = Date::Today() - 1;
	return 1;
    }

    // "<int>/<int>[/<int>]"
    if ((list[index+0].ttype == INT_TOKEN) &&
	(list[index+1].ttype == SLASH_TOKEN) &&
	(list[index+2].ttype == INT_TOKEN)) {

	int m = list[index+0].tvalue;
	int d = list[index+2].tvalue;
	if (m > 12) {
	    // European format
	    int tmp = m;
	    m = d;
	    d = tmp;
	}

	int count;
	int y;
	if ((list[index+3].ttype == SLASH_TOKEN) &&
	    (list[index+4].ttype == INT_TOKEN)) {
	    // Found year as well
	    y = list[index+4].tvalue;
	    count = 5;
	}
	else {
	    // Only month and year
	    y = Date::Today().GetYear();
	    count = 3;
	}

	if (build_date(m, d, y, result))
	    return count;
    }

    // "<month> <int> [<time>] [<int>]"
    if ((list[index+0].ttype == MONTH_TOKEN) &&
	(list[index+1].ttype == INT_TOKEN)) {

	int m = list[index+0].tvalue;
	int d = list[index+1].tvalue;
	int y = Date::Today().GetYear();
	int c = 2;

	// Skip time specification (if any)
	int seconds;
	c += match_timeofday(list, index+c, seconds);

	// Handle year specification (if any)
	if (list[index+c].ttype == INT_TOKEN) {
	    y = list[index+c].tvalue;
	    c++;
	}

	if (build_date(m, d, y, result))
	    return c;
    }

    // "<int> <month> [<time>] [<int>]"
    if ((list[index+0].ttype == INT_TOKEN) &&
	(list[index+1].ttype == MONTH_TOKEN)) {

	int m = list[index+1].tvalue;
	int d = list[index+0].tvalue;
	int y = Date::Today().GetYear();
	int c = 2;

	// Skip time specification (if any)
	int seconds;
	c += match_timeofday(list, index+c, seconds);

	// Handle year specification (if any)
	if (list[index+c].ttype == INT_TOKEN) {
	    y = list[index+c].tvalue;
	    c++;
	}

	if (build_date(m, d, y, result))
	    return c;
    }

    return 0;
}

// Exported routines

int find_date(char const* string, Date& result, int& start, int& length) {
    Tokens list;
    parse_tokens(string, list);

    int index = 0;		// Index into "list"
    int pos = 0;		// Number of chars in "list[0..index-1]".
    while (index < list.size()) {
	Date d;
	int count = match_date(list, index, d);
	if (count == 0) {
	    // No date here, keep searching
	    pos += list[index].len;
	    index++;
	    continue;
	}

	// Found date at "index"
	result = d;
	start = pos;
	length = 0;
	for (int i = index; i < index + count; i++)
	    length += list[i].len;

	return 1;
    }

    return 0;
}

int find_timeofday(char const* string, int& result, int& start, int& length) {
    Tokens list;
    parse_tokens(string, list);

    int index = 0;		// Index into "list"
    int pos = 0;		// Number of chars in "list[0..index-1]".
    while (index < list.size()) {
	int seconds;
	int count = match_timeofday(list, index, seconds);
	if (count == 0) {
	    // No time here, keep searching
	    pos += list[index].len;
	    index++;
	    continue;
	}

	// Found time at "index"
	result = seconds;
	start = pos;
	length = 0;
	for (int i = index; i < index + count; i++)
	    length += list[i].len;

	return 1;
    }

    return 0;
}

int find_timerange(char const* string,
		   int& startresult, int& finishresult,
		   int& start, int& length)
{
    Tokens list;
    parse_tokens(string, list);

    int index = 0;		// Index into "list"
    int pos = 0;		// Number of chars in "list[0..index-1]".
    while (index < list.size()) {
	int stseconds, fiseconds;
	int count = match_timerange(list, index, stseconds, fiseconds);
	if (count == 0) {
	    // No time here, keep searching
	    pos += list[index].len;
	    index++;
	    continue;
	}

	// Found time at "index"
	startresult = stseconds;
	finishresult = fiseconds;
	start = pos;
	length = 0;
	for (int i = index; i < index + count; i++)
	    length += list[i].len;

	return 1;
    }

    return 0;
}