Monday, June 01, 2009
And here I thought dating was easy …
I'm building on the work of indexing my filesystem by indexing all of my email. I have a ton of it spread across various directories and when ever I have to search for something (such as the time I flamed an entire department at FAU on a public mailing list—ah, those were the days), it's a long drawn out ordeal to find it.
Initial stab at the problem is to just index a few email headers, like
From:, To: (and the related Cc:),
Date: and Subject:—the primary headers one would be
interested in.
I decided to tackle one of the harder fields to process first—
From:. While the format is specified in RFC-822 and RFC-2822, there's still quite
a bit of variance in the format to be annoying.
I was able to squish 23 different formats into four cases:
- email address and real name aren't delimited, in which case, the only thing to parse is the email address;
- email isn't delimited, but the real name is (between parentheses, or quotes), so extract the real name from between the delimeters, and anything that isn't delimited is the email address;
- email is delimited (between angle brackets or square brackets), but the real name isn't, so extract the email address, and anything that isn't delimited is the real name;
- both the email address and real name are delimited, so it's trivial to extract both.
Then, I decided to parse the Date: header. Now, this
is specified, quite plainly:
5. DATE AND TIME SPECIFICATION
5.1. SYNTAX
date-time = [ day "," ] date time ; dd mm yy
; hh:mm:ss zzz
day = "Mon" / "Tue" / "Wed" / "Thu"
/ "Fri" / "Sat" / "Sun"
date = 1*2DIGIT month 2DIGIT ; day month year
; e.g. 20 Jun 82
month = "Jan" / "Feb" / "Mar" / "Apr"
/ "May" / "Jun" / "Jul" / "Aug"
/ "Sep" / "Oct" / "Nov" / "Dec"
time = hour zone ; ANSI and Military
hour = 2DIGIT ":" 2DIGIT [":" 2DIGIT]
; 00:00:00 - 23:59:59
zone = "UT" / "GMT" ; Universal Time
; North American : UT
/ "EST" / "EDT" ; Eastern: - 5/ - 4
/ "CST" / "CDT" ; Central: - 6/ - 5
/ "MST" / "MDT" ; Mountain: - 7/ - 6
/ "PST" / "PDT" ; Pacific: - 8/ - 7
/ 1ALPHA ; Military: Z = UT;
; A:-1; (J not used)
; M:-12; N:+1; Y:+12
/ ( ("+" / "-") 4DIGIT ) ; Local differential
; hours+min. (HHMM)
STANDARD FOR ARPA INTERNET TEXT MESSAGES, § 5.1
Okay, clear if you're into such things. And from the most recent specification:
date-time = [ day-of-week "," ] date FWS time [CFWS]
day-of-week = ([FWS] day-name) / obs-day-of-week
day-name = "Mon" / "Tue" / "Wed" / "Thu" /
"Fri" / "Sat" / "Sun"
date = day month year
year = 4*DIGIT / obs-year
month = (FWS month-name FWS) / obs-month
month-name = "Jan" / "Feb" / "Mar" / "Apr" /
"May" / "Jun" / "Jul" / "Aug" /
"Sep" / "Oct" / "Nov" / "Dec"
day = ([FWS] 1*2DIGIT) / obs-day
time = time-of-day FWS zone
time-of-day = hour ":" minute [ ":" second ]
hour = 2DIGIT / obs-hour
minute = 2DIGIT / obs-minute
second = 2DIGIT / obs-second
zone = (( "+" / "-" ) 4DIGIT) / obs-zone
RFC-2822: Internet Message Format, § 3.3
Really, the only thing this does is mandate that the year be four digits long, moves to a numeric-only timezone format and clarifies a bit where white space appears, but otherwise, is pretty much the same as the older spec.
So, if I ignore the timezone for now (because the Standard C library has such piss-poor support for it, but that's a rant for another time), the only real issue is handling two or four digit years.
And in poking around the man pages for the various Standard C library
routines, I came across strptime(), which is the functional
opposite of strftime()—instead of converting the time to a human
representation, it'll take a human representation and convert it to a time
value. It isn't a Standard C call, but hey, why not use it for now?
And it appears that the two-digit/four-digit year isn't a problem for
strptime():
When a century is not otherwise specified, values in the range [69,99] shall refer to years 1969 to 1999 inclusive, and values in the range [00,68] shall refer to years 2000 to 2068 inclusive; leading zeros shall be permitted but shall not be required.
man page for
strptime()
Sounds perfect!
Only it blew up when it encounted Wen, 2 Mar 2005 01:39:42
+0000.
Sigh.
Okay, make sure I start parsing past the optional day of the week. It then
blew up on Sat Mar 5 18:58:36 2005.
What the—? That's not even a standard format! And then there was
Wed,19 十二月 2001 20:23:05
(I added the question marks because I
can't determine the character set that was used for the month—there's
nothing in that particular email that even hints what language it
might be I found out which
language---Chinese. Figures).
And let's not forget 9/8/99 1:01:12 AM Pacific Daylight Time
(lovely) or Fri Jun 28 10:07:44 PDT 2002 or even
Wed 8-Jan-2003 08:24:20.
Oh, and we mustn't forget Tue, 23 May 100 22:18:56 -0400.
Double sigh.
I found it amazing—one of the more strictly defined fields in an email and yet there still was an amazing amount of garbage to be found (although to be fair, these anomalies account for less than one per cent of all the emails scanned, but when you have thousands of emails, it can still add up).
(And one more interesting note—I did not see one email use the military time zone format.)
