/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.pdfbox.util;
import java.text.ParsePosition;
import java.text.SimpleDateFormat;
import java.util.Calendar;
import java.util.Date;
import java.util.GregorianCalendar;
import java.util.Locale;
import java.util.SimpleTimeZone;
import java.util.TimeZone;
import org.apache.pdfbox.cos.COSString;
/*
* Date format is described in PDF Reference 1.7 section 3.8.2
* (www.adobe.com/devnet/acrobat/pdfs/pdf_reference_1-7.pdf)
* and also in PDF 32000-1:2008
* (http://www.adobe.com/devnet/acrobat/pdfs/PDF32000_2008.pdf))
* although the latter inexplicably omits the trailing apostrophe.
*
* The interpretation of dates without timezones is unclear.
* The code below assumes that such dates are in UTC+00 (aka GMT).
* This is in keeping with the PDF Reference's assertion that:
* numerical fields default to zero values.
* However, the Reference does go on to make the cryptic remark:
* If no UT information is specified, the relationship of the specified
* time to UT is considered to be unknown. Whether or not the time
* zone is known, the rest of the date should be specified in local time.
* I understand this to refer to _creating_ a pdf date value. That is,
* code that can get the wall clock time and cannot get the timezone
* should write the wall clock time with a time zone of zero.
* When _parsing_ a PDF date, the statement talks about "the rest of the date"
* being local time, thus explicitly excluding the use of the local time
* for the time zone.
*/
/**
* Converts dates to strings and back using the PDF date standard
* in section 3.8.2 of PDF Reference 1.7.
*
* @author Ben Litchfield
* @author Fred Hansen
*
* TODO Move members of this class elsewhere for shared use in pdfbox and xmpbox.
*/
public final class DateConverter
{
private DateConverter()
{
}
// milliseconds/1000 = seconds; seconds / 60 = minutes; minutes/60 = hours
private static final int MINUTES_PER_HOUR = 60;
private static final int SECONDS_PER_MINUTE = 60;
private static final int MILLIS_PER_MINUTE = SECONDS_PER_MINUTE*1000;
private static final int MILLIS_PER_HOUR = MINUTES_PER_HOUR * MILLIS_PER_MINUTE;
private static final int HALF_DAY = 12 * MINUTES_PER_HOUR * MILLIS_PER_MINUTE, DAY = 2*HALF_DAY;
/*
* The Date format is supposed to be the PDF_DATE_FORMAT, but other
* forms appear. These lists offer alternatives to be tried
* if parseBigEndianDate fails.
*
* The time zone offset generally trails the date string, so it is processed
* separately with parseTZoffset. (This does not preclude having time
* zones in the elements below; one does.)
*
* Alas, SimpleDateFormat is badly non-reentrant -- it modifies its
* calendar field (PDFBox-402), so these lists are strings to create
* SimpleDate format as needed.
*
* Some past entries have been elided because they duplicate existing
* entries. See the API for SimpleDateFormat, which says
* "For parsing, the number of pattern letters is ignored
* unless it's needed to separate two adjacent fields."
*
* toCalendar(String, String[]) tests to see that the entire input text
* has been consumed. Therefore the ordering of formats is important.
* If one format begins with the entirety of another, the longer
* must precede the other in the list.
*
* HH is for 0-23 hours and hh for 1-12 hours; an "a" field must follow "hh"
* Where year is yy, four digit years are accepted
* and two digit years are converted to four digits in the range
* [thisyear-79...thisyear+20]
*/
private static final String[] ALPHA_START_FORMATS =
{
"EEEE, dd MMM yy hh:mm:ss a",
"EEEE, MMM dd, yy hh:mm:ss a",
"EEEE, MMM dd, yy 'at' hh:mma", // Acrobat Net Distiller 1.0 for Windows
"EEEE, MMM dd, yy", // Acrobat Distiller 1.0.2 for Macintosh && PDFBOX-465
"EEEE MMM dd, yy HH:mm:ss", // ECMP5
"EEEE MMM dd HH:mm:ss z yy", // GNU Ghostscript 7.0.7
"EEEE MMM dd HH:mm:ss yy", // GNU Ghostscript 7.0.7 variant
};
private static final String[] DIGIT_START_FORMATS =
{
"dd MMM yy HH:mm:ss", // for 26 May 2000 11:25:00
"dd MMM yy HH:mm", // for 26 May 2000 11:25
"yyyy MMM d", // ambiguity resolved only by omitting time
"yyyymmddhh:mm:ss", // test case "200712172:2:3"
"H:m M/d/yy", // test case "9:47 5/12/2008"
"M/d/yy HH:mm:ss",
"M/d/yy HH:mm",
"M/d/yy",
// proposed rule that is unreachable due to "dd MMM yy HH:mm:ss"
// "yyyy MMM d HH:mm:ss",
// rules made unreachable by "M/d/yy HH:mm:ss" "M/d/yy HH:mm" "M/d/yy",
// (incoming digit strings do not mark themselves as y, m, or d!)
// "d/MM/yyyy HH:mm:ss", // PDFBOX-164 and PDFBOX-170
// "M/dd/yyyy hh:mm:ss",
// "MM/d/yyyy hh:mm:ss",
// "M/d/yyyy HH:mm:ss",
// "M/dd/yyyy",
// "MM/d/yyyy",
// "M/d/yyyy",
// "M/d/yyyy HH:mm:ss",
// "M/d/yy HH:mm:ss",
// subsumed by big-endian parse
// "yyyy-MM-dd'T'HH:mm:ss",
// "yyyy-MM-dd'T'HH:mm:ss",
// "yyyymmdd hh:mm:ss",
// "yyyymmdd",
// "yyyymmddX''00''", // covers 24 cases
// (orignally the above ended with '+00''00''';
// the first apostrophe quoted the plus,
// '' mapped to a single ', and the ''' was invalid)
};
/**
* Converts a Calendar to a string formatted as:
* D:yyyyMMddHHmmss#hh'mm' where # is Z, +, or -.
*
* @param cal The date to convert to a string. May be null.
* The DST_OFFSET is included when computing the output time zone.
*
* @return The date as a String to be used in a PDF document,
* or null if the cal value is null
*/
public static String toString(Calendar cal)
{
if (cal == null)
{
return null;
}
String offset = formatTZoffset(cal.get(Calendar.ZONE_OFFSET) +
cal.get(Calendar.DST_OFFSET), "'");
return String.format("D:"
+ "%1$4tY%1$2tm%1$2td" // yyyyMMdd
+ "%1$2tH%1$2tM%1$2tS" // HHmmss
+ "%2$s" // time zone
+ "'", // trailing apostrophe
cal, offset);
}
/**
* Converts the date to ISO 8601 string format:
* yyyy-mm-ddThh:MM:ss#hh:mm (where '#" is '+' or '-').
*
* @param cal The date to convert. Must not be null.
* The DST_OFFSET is included in the output value.
*
* @return The date represented as an ISO 8601 string.
*/
public static String toISO8601(Calendar cal)
{
String offset = formatTZoffset(cal.get(Calendar.ZONE_OFFSET) +
cal.get(Calendar.DST_OFFSET), ":");
return String.format(
"%1$4tY" // yyyy
+ "-%1$2tm" // -mm (%tm adds one to cal month value)
+ "-%1$2td" // -dd (%tm adds one to cal month value)
+ "T" // T
+ "%1$2tH:%1$2tM:%1$2tS" // HHmmss
+ "%2$s", // time zone
cal, offset);
}
/*
* Constrain a timezone offset to the range [-11:59 thru +11:59].
* by adding or subtracting multiples of a full day.
*/
private static int restrainTZoffset(long proposedOffset)
{
proposedOffset = ((proposedOffset + HALF_DAY) % DAY + DAY) % DAY;
// 0 <= proposedOffset < DAY
proposedOffset = (proposedOffset - HALF_DAY) % HALF_DAY;
// -HALF_DAY < proposedOffset < HALF_DAY
return (int)proposedOffset;
}
/*
* Formats a time zone offset as #hh^mm
* where # is + or -, hh is hours, ^ is a separator, and mm is minutes.
* Any separator may be specified by the second argument;
* the usual values are ":" (ISO 8601), "" (RFC 822), and "'" (PDF).
* The returned value is constrained to the range -11:59 ... 11:59.
* For offset of 0 millis, the String returned is "+00^00", never "Z".
* To get a "general" offset in form GMT#hh:mm, write
* "GMT"+DateConverter.formatTZoffset(offset, ":");
*
* Take thought in choosing the source for the millis value.
* It can come from calendarValue.getTimeZone() or from
* calendarValue.get(Calendar.ZONE_OFFSET). If a TimeZone was created
* from a valid time zone ID, then it may have a daylight savings rule.
* (As of July 4, 2013, the data base at http://www.iana.org/time-zones
* recognized 629 time zone regions. But a TimeZone created as
* new SimpleTimeZone(millisOffset, "ID"),
* will not have a daylight savings rule. (Not even if there is a
* known time zone with the given ID. To get the TimeZone named "xDT"
* with its DST rule, use an ID of EST5EDT, CST6CDT, MST7MDT, or PST8PDT.
*
* When parsing PDF dates, the incoming values DOES NOT have a TIMEZONE value.
* At most it has an OFFSET value like -04'00'. It is generally impossible to
* determine what TIMEZONE corresponds to a given OFFSET. If the date is
* in the summer when daylight savings is in effect, an offset of -0400
* might correspond to any one of the 38 regions (of 53) with standard time
* offset -0400 and no daylight saving. Or it might correspond to
* any one of the 31 regions (out of 43) that observe daylight savings
* and have standard time offset of -0500.
*
* If a Calendar has not been assigned a TimeZone with setTimeZone(),
* it will have by default the local TIMEZONE, not just the OFFSET. In the
* USA, this TimeZone will have a daylight savings rule.
*
* The offset assigned with calVal.set(Calendar.ZONE_OFFSET) differs
* from the offset in the TimeZone set by Calendar.setTimeZone(). Example:
* Suppose my local TimeZone is America/New_York. It has an offset of -05'00'.
* And suppose I set a GregorianCalendar's ZONE_OFFSET to -07'00'
* calVal = new GregorianCalendar(); // TimeZone is the local default
* calVal.set(Calendar.ZONE_OFFSET, -7* MILLIS_PER_HOUR);
* Four different offsets can be computed from calVal:
* calVal.get(Calendar.ZONE_OFFSET) => -07:00
* calVal.get(Calendar.ZONE_OFFSET) + calVal.get(Calendar.DST_OFFSET) => -06:00
* calVal.getTimeZone().getRawOffset() => -05:00
* calVal.getTimeZone().getOffset(calVal.getTimeInMillis()) => -04:00
*
* Which is correct??? I dunno, though setTimeZone() does seem to affect
* ZONE_OFFSET, and not vice versa. One cannot even test whether TimeZone
* or ZONE_OFFSET has been set; both have been set by initialization code.
* TimeZone is initialized to the local default time zone
* and ZONE_OFFSET is set from it.
*
* My choice in this DateConverter class has been to set the
* initial TimeZone of a GregorianCalendar to GMT. Thereafter
* the TimeZone is modified with {@link #adjustTimeZoneNicely}.
*
* package-private for testing
*/
static String formatTZoffset(long millis, String sep)
{
SimpleDateFormat sdf = new SimpleDateFormat("Z"); // #hhmm
sdf.setTimeZone(new SimpleTimeZone(restrainTZoffset(millis),"unknown"));
String tz = sdf.format(new Date());
return tz.substring(0,3) + sep + tz.substring(3);
}
/*
* Parses an integer from a string, starting at and advancing a ParsePosition.
* Returns The integer that was at the given parse position, or the remedy value
* if no digits were found.
*
* The ParsePosition will be incremented by the number of digits found, but no
* more than maxlen. That is, the ParsePosition will advance across at most
* maxlen initial digits in text. The error index is ignored and unchanged.
*
* maxlen is the maximum length of the integer to parse, usually 2, but 4 for
* year fields. If the field of length maxlen begins with a digit, but contains
* a non-digit, no error is signaled and the integer value is returned.
*/
private static int parseTimeField(String text, ParsePosition where, int maxlen, int remedy)
{
if (text == null)
{
return remedy;
}
// it would seem that DecimalFormat.parse() would be simpler;
// but that class blithely ignores setMaximumIntegerDigits
int retval = 0;
int index = where.getIndex();
int limit = index + Math.min(maxlen, text.length()-index);
for (; index < limit; index++)
{
// convert digit to integer
int cval = text.charAt(index) - '0';
// test to see if we got a digit
if (cval < 0 || cval > 9)
{
// no digit at index
break;
}
// append the digit to the return value
retval = retval * 10 + cval;
}
if (index == where.getIndex())
{
return remedy;
}
where.setIndex(index);
return retval;
}
/*
* Advances the ParsePosition past any and all the characters that match
* those in the optionals list. In particular, a space will skip all spaces.
*
* The start value is incremented by the number of optionals found. The error
* index is ignored and unchanged.
*
* Returns the last non-space character passed over (even if space is not in
* the optionals list.)
*/
private static char skipOptionals(String text, ParsePosition where, String optionals)
{
char retval = ' ', currch;
while (text != null && where.getIndex() < text.length() &&
optionals.indexOf((currch = text.charAt(where.getIndex()))) >= 0)
{
retval = (currch != ' ') ? currch : retval;
where.setIndex(where.getIndex() + 1);
}
return retval;
}
/*
* If the victim string is at the given position in the text, this method
* advances the position past that string.
*
* `where` is the initial position to look at. After return, this will have
* been incremented by the length of the victim if it was found. The error
* index is ignored and unchanged.
*/
private static boolean skipString(String text, String victim, ParsePosition where)
{
if (text.startsWith(victim, where.getIndex()))
{
where.setIndex(where.getIndex()+victim.length());
return true;
}
return false;
}
/*
* Construct a new GregorianCalendar and set defaults.
* Locale is ENGLISH.
* TimeZone is "UTC" (zero offset and no DST).
* Parsing is NOT lenient. Milliseconds are zero.
*
* package-private for testing
*/
static GregorianCalendar newGreg()
{
GregorianCalendar retCal = new GregorianCalendar(Locale.ENGLISH);
retCal.setTimeZone(new SimpleTimeZone(0, "UTC"));
retCal.setLenient(false);
retCal.set(Calendar.MILLISECOND, 0);
return retCal;
}
/*
* Install a TimeZone on a GregorianCalendar without changing the
* hours value. A plain GregorianCalendat.setTimeZone()
* adjusts the Calendar.HOUR value to compensate. This is *BAD*
* (not to say *EVIL*) when we have already set the time.
*/
private static void adjustTimeZoneNicely(GregorianCalendar cal, TimeZone tz)
{
cal.setTimeZone(tz);
int offset = (cal.get(Calendar.ZONE_OFFSET) + cal.get(Calendar.DST_OFFSET)) /
MILLIS_PER_MINUTE;
cal.add(Calendar.MINUTE, -offset);
}
/*
* Parses the end of a date string for a time zone and, if one is found,
* sets the time zone of the GregorianCalendar. Otherwise the calendar
* time zone is unchanged.
*
* The text is parsed as
* (Z|GMT|UTC)? [+- ]* h [': ]? m '?
* where the leading String is optional, h is two digits by default,
* but may be a single digit if followed by one of space, apostrophe,
* colon, or the end of string. Similarly, m is one or two digits.
* This scheme accepts the format of PDF, RFC 822, and ISO8601.
* If none of these applies (as for a time zone name), we try
* TimeZone.getTimeZone().
*
* Scanning begins at where.index. After success, the returned index
* is that of the next character after the recognized string.
*
* package-private for testing
*/
static boolean parseTZoffset(String text, GregorianCalendar cal,
ParsePosition initialWhere)
{
ParsePosition where = new ParsePosition(initialWhere.getIndex());
TimeZone tz = new SimpleTimeZone(0, "GMT");
int tzHours, tzMin;
char sign = skipOptionals(text, where, "Z+- ");
boolean hadGMT = (sign == 'Z' || skipString(text, "GMT", where) ||
skipString(text, "UTC", where));
sign = (!hadGMT) ? sign : skipOptionals(text, where, "+- ");
tzHours = parseTimeField(text, where, 2, -999);
skipOptionals(text, where, "\': ");
tzMin = parseTimeField(text, where, 2, 0);
skipOptionals(text, where, "\' ");
if (tzHours != -999)
{
// we parsed a time zone in default format
int hrSign = (sign == '-' ? -1 : 1);
tz.setRawOffset(restrainTZoffset(hrSign * (tzHours * MILLIS_PER_HOUR + tzMin *
MILLIS_PER_MINUTE)));
tz.setID("unknown");
}
else if ( ! hadGMT)
{
// try to process as a name; "GMT" or "UTC" has already been processed
String tzText = text.substring(initialWhere.getIndex()).trim();
tz = TimeZone.getTimeZone(tzText);
// getTimeZone returns "GMT" for unknown ids
if ("GMT".equals(tz.getID()))
{
// no timezone in text, cal amd initialWhere are unchanged
return false;
}
else
{
// we got a tz by name; use it
where.setIndex(text.length());
}
}
adjustTimeZoneNicely(cal, tz);
initialWhere.setIndex(where.getIndex());
return true;
}
/*
* Parses a big-endian date: year month day hour min sec.
* The year must be four digits. Other fields may be adjacent
* and delimited by length or they may follow appropriate delimiters.
* year [ -/]* month [ -/]* dayofmonth [ T]* hour [:] min [:] sec [.secFraction]
* If any numeric field is omitted, all following fields must also be omitted.
* No time zone is processed.
*
* Ambiguous dates can produce unexpected results. For example:
* 1970 12 23:08 will parse as 1970 December 23 00:08:00
*
* The parse begins at `where, on return the index
* is advanced to just beyond the last character processed.
* The error index is ignored and unchanged.
*/
private static GregorianCalendar parseBigEndianDate(String text,
ParsePosition initialWhere)
{
ParsePosition where = new ParsePosition(initialWhere.getIndex());
int year = parseTimeField(text, where, 4, 0);
if (where.getIndex() != 4 + initialWhere.getIndex())
{
return null;
}
skipOptionals(text, where, "/- ");
int month = parseTimeField(text, where, 2, 1) - 1; // Calendar months are 0...11
skipOptionals(text, where, "/- ");
int day = parseTimeField(text, where, 2, 1);
skipOptionals(text, where, " T");
int hour = parseTimeField(text, where, 2, 0);
skipOptionals(text, where, ": ");
int minute = parseTimeField(text, where, 2, 0);
skipOptionals(text, where, ": ");
int second = parseTimeField(text, where, 2, 0);
char nextC = skipOptionals(text, where, ".");
if (nextC == '.')
{
// fractions of a second: skip upto 19 digits
parseTimeField(text, where, 19, 0);
}
GregorianCalendar dest = newGreg();
try
{
dest.set(year, month, day, hour, minute, second);
// trigger limit tests
dest.getTimeInMillis();
}
catch (IllegalArgumentException ill)
{
return null;
}
initialWhere.setIndex(where.getIndex());
skipOptionals(text, initialWhere, " ");
// dest has at least a year value
return dest;
}
/*
* See if text can be parsed as a date according to any of a list of
* formats. The time zone may be included as part of the format, or
* omitted in favor of later testing for a trailing time zone.
*
* The parse starts at `where`, upon return it will have been
* incremented to refer to the next non-space character after the date.
* If no date was found, the value is unchanged.
* The error index is ignored and unchanged.
*
* If there is a failure to find a date, or the GregorianCalendar
* for the date that was found. Unless a time zone was
* part of the format, the time zone will be GMT+0
*/
private static GregorianCalendar parseSimpleDate(String text, String[] fmts,
ParsePosition initialWhere)
{
for(String fmt : fmts)
{
ParsePosition where = new ParsePosition(initialWhere.getIndex());
SimpleDateFormat sdf = new SimpleDateFormat(fmt, Locale.ENGLISH);
GregorianCalendar retCal = newGreg();
sdf.setCalendar(retCal);
if (sdf.parse(text, where) != null)
{
initialWhere.setIndex(where.getIndex());
skipOptionals(text, initialWhere, " ");
return retCal;
}
}
return null;
}
/*
* Parses a String to see if it begins with a date, and if so,
* returns that date. The date must be strictly correct--no
* field may exceed the appropriate limit.
* (That is, the Calendar has setLenient(false).)
* Skips initial spaces, but does NOT check for "D:"
*
* The scan first tries parseBigEndianDate and parseTZoffset
* and then tries parseSimpleDate with appropriate formats,
* again followed by parseTZoffset. If at any stage the entire
* text is consumed, that date value is returned immediately.
* Otherwise the date that consumes the longest initial part
* of the text is returned.
*
* - PDF format dates are among those recognized by parseBigEndianDate.
* - The formats tried are alphaStartFormats or digitStartFormat and
* any listed in the value of moreFmts.
*/
private static Calendar parseDate(String text, ParsePosition initialWhere)
{
if (text == null || text.isEmpty())
{
return null;
}
// remember longestr date string
int longestLen = -999999;
// theorem: the above value will never be used
// proof: longestLen is only used if longestDate is not null
GregorianCalendar longestDate = null; // null says no date found yet
int whereLen; // tempcopy of where.getIndex()
ParsePosition where = new ParsePosition(initialWhere.getIndex());
// check for null (throws exception) and trim off surrounding spaces
skipOptionals(text, where, " ");
int startPosition = where.getIndex();
// try big-endian parse
GregorianCalendar retCal = parseBigEndianDate(text, where);
// check for success and a timezone
if (retCal != null && (where.getIndex() == text.length() ||
parseTZoffset(text, retCal, where)))
{
// if text is fully consumed, return the date else remember it and its length
whereLen = where.getIndex();
if (whereLen == text.length())
{
initialWhere.setIndex(whereLen);
return retCal;
}
longestLen = whereLen;
longestDate = retCal;
}
// try one of the sets of standard formats
where.setIndex(startPosition);
String [] formats
= Character.isDigit(text.charAt(startPosition))
? DIGIT_START_FORMATS
: ALPHA_START_FORMATS;
retCal = parseSimpleDate(text, formats, where);
// check for success and a timezone
if (retCal != null &&
(where.getIndex() == text.length() ||
parseTZoffset(text, retCal, where)))
{
// if text is fully consumed, return the date else remember it and its length
whereLen = where.getIndex();
if (whereLen == text.length())
{
initialWhere.setIndex(whereLen);
return retCal;
}
if (whereLen > longestLen)
{
longestLen = whereLen;
longestDate = retCal;
}
}
if (longestDate != null)
{
initialWhere.setIndex(longestLen);
return longestDate;
}
return retCal;
}
/**
* Returns the Calendar for a given COS string containing a date,
* or {@code null} if it cannot be parsed.
*
* The returned value will have 0 for DST_OFFSET.
*
* @param text A COS string containing a date.
* @return The Calendar that the text string represents, or {@code null} if it cannot be parsed.
*/
public static Calendar toCalendar(COSString text)
{
if (text == null)
{
return null;
}
return toCalendar(text.getString());
}
/**
* Returns the Calendar for a given string containing a date,
* or {@code null} if it cannot be parsed.
*
* The returned value will have 0 for DST_OFFSET.
*
* @param text A COS string containing a date.
* @return The Calendar that the text string represents, or {@code null} if it cannot be parsed.
*/
public static Calendar toCalendar(String text)
{
if (text == null || text.trim().isEmpty())
{
return null;
}
ParsePosition where = new ParsePosition(0);
skipOptionals(text, where, " ");
skipString(text, "D:", where);
Calendar calendar = parseDate(text, where);
if (calendar == null || where.getIndex() != text.length())
{
// the date string is invalid
return null;
}
return calendar;
}
}