Date::Manip::Problems(User Contributed Perl DocumentatDate::Manip::Problems(3)NAMEDate::Manip::Problems - problems and bugs
KNOWN PROBLEMS
The following are not bugs in Date::Manip, but they may give some
people problems.
Memory leak
There is a known memory leak in perl related to named regexp
captures that directly affects Date::Manip . The leak is in all
versions of perl that I have tested starting at 5.10. I have
submitted a bug report (perlbug #78266). Until perl is fixed, if
you need to use Date::Manip in a persistent environment, you should
be aware of this.
Date::Manip 5.xx is not susceptible, so using it may be a feasible
workaround, but if you need accurate timezone handling, this isn't
possible.
The leak is fairly small. A crude estimate is 3 MB per 10,000 dates
parsed, so unless you're parsing hundreds of thousands, or millions
of dates, the leak probably won't be a problem. And if you're
parsing that many dates, the relatively slow Date::Manip may not be
the correct module for you to use anyway.
Unable to determine Time Zone
Please refer to the Date::Manip::TZ documentation for a discussion
of this problem.
Calculations appear to be off by an hour
Within days of releasing 6.00, I got the following report:
print DateCalc(ParseDate("2009/10/31"), "+1 day"), "\n";
print DateCalc(ParseDate("2009/11/01"), "+1 day"), "\n";
print DateCalc(ParseDate("2009/11/02"), "+1 day"), "\n";
produced:
2009110100:00:00
2009110123:00:00
2009110300:00:00
In other words, the second calculation appears to be off by one
hour. In actuality, the result is correct. In the user's time
zone, daylight saving time ended on Nov 1, and since "+ 1 day" is
equivalent to "+ 24 hours", the results are correct. In
America/New_York time (which has the same result), the calculation
means that:
2009-11-01 00:00:00 EDT + 24 hours =
2009-11-01 23:00:00 EST
which is correct.
Missing date formats
Due to the large number of date formats that Date::Manip CAN
process, people often assume that other formats that they want to
use should work as well, and when they don't, it comes as a
surprise.
With the much improved parsing of 6.00, many formats can be added
easily, though unless they are of general use, I'll probably
suggest that you use parse_format instead.
There is a class of formats that I will not add however.
I have frequently been asked to add formats such as "the 15th of
last month", or "Monday of next week". I will NOT add these date
formats to Date::Manip. Since I have received the request several
times, I decided to include my reasoning here.
Date::Manip can parse pretty much any static date format that I
could think of or find reference to. Dates such as "today", "Jan
12", or "2001-01-01" are all understood.
These are fairly limited however. Many very common date formats are
best thought of as a date plus a modification. For example,
"yesterday" is actually determined internally as "today" plus a
modification of "- 1 day". "2nd Sunday in June" is determined as
"June 1" modified to the 2nd Sunday.
As these types of formats were added over time, I quickly realized
that the number of possible date plus modification formats was
huge. The number of combinations has caused the parsing in
Date::Manip to be quite complex, and adding new formats
occasionally causes unexpected conflicts with other formats.
The first time I received a request similar to "the 15th of last
month", I intended to add it, but as I analyzed it to see what
changes needed to be made to support it, I realized that this
needed to be expressed as a date plus TWO modifications. In other
words, today modified to last month modified to the 15th day of the
month.
As bad as date plus modification formats are, a date plus TWO
modifications would be exponentially worse. On realizing that, I
made a firm decision that Date::Manip will NOT support this type of
format now, or at any time in the future. Although I apologize for
the inconvenience, I do not intend to change my position on this.
Date::Manip is slow
NOTE: The following section applies primarily to 5.xx. I'm doing a
lot of work to optimize Date::Manip and I will rewrite this section
to take this into account, and to provide performance suggestions.
It should be noted that initial tests show version 6.xx to be
around twice as fast as 5.xx (though still considerably slower than
some of the other modules).
Date::Manip is probably one of the slower Date/Time modules due to
the fact that it is huge and written entirely in perl.
Some things that will definitely help:
ISO-8601 dates are parsed first and fastest. Use them whenever
possible.
Avoid parsing dates that are referenced against the current time
(in 2 days, today at noon, etc.). These take a lot longer to
parse.
Business date calculations are extremely slow. You should consider
alternatives if possible (i.e. doing the calculation in exact mode
and then multiplying by 5/7). Who needs a business date more
accurate than "6 to 8 weeks" anyway, right :-)
RCS Control
If you try to put Date::Manip under RCS control, you are going to
have problems. Apparently, RCS replaces strings of the form
"$Date...$" with the current date. This form occurs all over in
Date::Manip. To prevent the RCS keyword expansion, checkout files
using:
co -ko
Since very few people will ever have a desire to do this (and I
don't use RCS), I have not worried about it, and I do not intend to
try to workaround this problem.
Using functions/methods which are not supported
There have been a handful of incidents of people using a function
from Date::Manip which were not documented in the manual.
Date::Manip consists of a large number of user functions which are
documented in the manual. These are designed to be used by other
programmers, and I will not make any backwards incompatible changes
in them unless there is a very compelling reason to do so, and in
that case, the change will be clearly documented in the
Date::Manip::Changes6 documentation for this module.
Date::Manip also includes a large number of functions which are NOT
documented. These are for internal use only. Please do not use
them! I can (and do) change their use, and even their name,
without notice, and without apology! Some of these internal
functions even have test scripts, but that is not a guarantee that
they will not change, nor is any support implied. I simply like to
run regression tests on as much of Date::Manip as possible.
As of the most recent versions of Date::Manip, all internal
functions have names that begin with an underscore (_). If you
choose to use them directly, it is quite possible that new versions
of Date::Manip will cause your programs to break due to a change in
how those functions work.
Any changes to internal functions will not be documented, and will
not be regarded by me as a backwards incompatibility. Nor will I
(as was requested in one instance) revert to a previous version of
the internal function.
If you feel that an internal function is of more general use, feel
free to contact me with an argument of why it should be "promoted".
I welcome suggestions and will definitely consider any such
request.
KNOWN COMPLAINTS
Date::Manip 6.xx has gotten some complaints (far more than 5.xx if the
truth be told), so I'd like to address a couple of them here. Perhaps
an understanding of why some of the changes were made will allay some
of the complaints. If not, people are always welcome to stick with the
5.xx release. I will continue to support the 5.xx releases for a couple
years (though I do NOT plan to add functionality to it).
These complaints come both from both the CPAN ratings site:
http://cpanratings.perl.org/dist/Date-Manip
and from personal email.
Requires perl 5.10
The single most controversial change made in 6.00 is that it now
required perl 5.10.0 or higher. Most of the negative feedback I've
received is due to this.
In the past, I've avoided using new features of perl in order to
allow Date::Manip to run on older versions of perl. Prior to perl
5.10, none of the new features would have had a major impact on how
Date::Manip was written, however that changed in 5.10.
One of the aspects of Date::Manip that has received the most
positive response is the ability to parse almost every conceivable
date format. Unfortunately, as I've added formats, the parsing
routine became more and more complicated, and maintaining it was
one of the least enjoyable aspect in maintaining Date::Manip . For
several years, I've been extremely reluctant to add new formats due
to the fact that too often, adding a new format broke other
formats.
As I was rewriting Date::Manip, I was looking for ways to improve
the parsing and to make maintaining it easier. Perl 5.10 provides
the feature "named capture buffers". Named capture buffers not only
improves the ease of maintaining the complex regular expressions
used by Date::Manip, it makes it dramatically easier to add
additional formats in a way that is much less likely to interfere
with other formats. The parsing in 6.00 is so much more robust,
extensible, and flexible, that it will make parser maintenance
possible for many years to come at a fraction of the effort and
risk.
It was too much to turn down. Hopefully, since 5.10 has been out
for some time now, this will not prohibit too many people from
using the new version of Date::Manip. I realize that there are many
people out there using older versions of perl who do not have the
option of upgrading perl. The decision to use 5.10 wasn't made
lightly... but I don't regret making it. I apologize to users who,
as a result, cannot use 6.00 . Hopefully in the future you'll be
able to benefit from the improvements in 6.00.
One complaint I've received is that this in some way makes
Date::Manip backwards incompatible, but this is not an accurate
complaint. Version 6.xx DOES include some backwards
incompatibilities (and these are covered in the
Date::Manip::Migration5to6 document), however in almost all cases,
these incompatibilities are with infrequently used features, or
workarounds are in place to allow deprecated features to continue
functioning for some period of time.
Though I have no data to confirm this, I suspect that 90% or more
of all scripts which were written with Date::Manip 5.xx will
continue to work unmodified with 6.xx (of course, you should still
refer to the migration document to see what features are deprecated
or changed to make sure that you don't need to modify your script
so that it will continue to work in the future). Even with scripts
that need to be changed, those changes should be trivial.
So, Date::Manip 6.xx is almost entirely backward compatible with
5.xx (to the extent that you would expect any major version release
to be compatible with a previous major version).
The change is only in the requirements necessary to get Date::Manip
6.xx to run.
Obviously, it's not reasonable to say that Date::Manip should never
be allowed to switch minimum perl versions. At some point, you have
to let go of an old version if you want to make use of the features
of the newer version. The question is, did I jump to 5.10 too fast?
The complaints I see on the CPAN ratings complain that I no longer
support perl 5.6 and perl 5.8.
With respect to 5.6, perl 5.6 was released in March of 2000 (that's
before Windows XP which was released in 2001). To be honest, I
don't really feel much sympathy for this complaint. Software that
is 9 years old is ANCIENT. People may choose to use it... but
please don't complain when new software comes out that doesn't
support it.
The argument for perl 5.8 is much more compelling. Although 5.8 was
released quite some time ago (July of 2002), there were no major
perl releases until 5.10 came out in December of 2007, so 5.8 was
state-of-the art as little as 2 years prior to the release of
Date::Manip 6.xx.
I agree completely with the argument that abandoning 5.8 only 2
years after it was the current version is too soon. For that
reason, I will continue to support the Date::Manip 5.xx releases
for some years to come. I don't know exactly how long I'll continue
to support them, but it'll be at least 2-3 years. Once perl 5.10 is
5 years old, I'll be much more likely to drop support for the 5.xx
releases, but I DO want to make use of the features of 5.10 for
future development. They make development so much easier, and the
parsing so much more robust (something I've wanted for years), that
I'm not willing to give up the advantages of 5.10.
But the next complaint is relevant.
Automatic installs break
A more important problem is that versions 6.01 through 6.07 broke
automatic installs for older perl installations. If you try to
install Date::Manip using the automatic tools (cpan/cpanp), they
will look for the most recent version. If you are using a version
of perl older than 5.10, this fails, and rather than looking for an
older version, the tool simply reports a failure in installing
Date::Manip. Technically, the problem is not due to Date::Manip
itself, but is a result of how perl modules are currently managed.
However, since Date::Manip is managed by then, it's important to
avoid causing this type of problem (which I clearly failed to do).
As of Date::Manip 6.10, this problem should no longer occur.
Starting with version 6.10, both the 5.xx and 6.xx versions of
Date::Manip have been combined into a single distribution (so
Date-Manip-6.10 contained both Date::Manip 6.10 and Date::Manip
5.57). From Date::Manip 6.10 to 6.13, the perl version was
determined at install time and either the 5.xx or 6.xx version was
installed. From Date::Manip 6.14 on, both versions are installed,
and at run time, the correct version will be chosen (and if you're
running a recent version of perl, you can select to run the old or
new version).
All future version (for as long as 5.xx is supported) will include
both the most current 5.xx and 6.xx releases of Date::Manip. In
this way, automatic install tools will be able to install
Date::Manip regardless of which version of perl you are running.
Too many modules
One minor complaint is that there are too many files. One person
specifically objects to the fact that there are over 470 modules
covering non-minute offsets. This complaint is (IMO) absurd.
Date::Manip supports ALL historical time zones, including those
with non-minute offsets, and so there will be information for those
time zones, even though they are not currently in use.
I could of course store all of the information in one big module,
but that means that you have to load all of that data every time
you use Date::Manip, and I find that to be a very poor solution.
Instead, storing the information in a per-time zone and per-offset
manner dramatically decreases the amount of data that has to be
loaded.
While it is true that Date::Manip includes over 900 modules for all
of the time zone information, most implementations of time zone
handling also choose to break up the data into a large number of
files.
My linux distribution (openSuSE 11.2 at the time of writing this)
uses the standard zoneinfo database, and at this point, there are
over 1700 files included in /usr/share/zoneinfo (though it does
appear that there is some duplication of information). Current
versions of RedHat also use over 1700 files, so Date::Manip isn't
treating the time zone problem in a new or unreasonable way.
Objects are large
One complaint that was put on the CPAN ratings site was that the OO
interface is "a dud" due to the size of it's objects. The complaint
is that a Date::Manip::Date object is 115K while it should
(according to the complaint) only require that you need to save the
seconds from the epoch, zone, and a couple other pieces of
information, all of which could probably be stored in 100 bytes or
less.
Date::Manip is very configurable, and contains a great deal of
information which could theoretically be calculated on the fly, but
which would greatly reduce it's performance. Instead, the data is
cached, and since the data is virtually all potentially object
specific, it has to be somehow linked to the object.
For example, Date::Manip allows you to parse dates in several
languages. Each language has a large number of regular expressions
which are used to to the actual parsing. Instead of recreating
these regular expressions each time they are needed, they are
created once and stored in an object (specifically, a
Date::Manip::Base object).
Similarly, a description of the time zones are stored in a second
object (a Date::Manip::TZ object).
The size of the Date::Manip::Base object is almost 15K. The size of
the Date::Manip::TZ object is 100K. That may seem excessive, but
you have to remember that there are almost 500 time zones, and they
have to be indexed by name, alias, abbreviation, and offset, and by
the time you do this, it does take a fair bit of space. The size
of the actual Date::Manip::Date object is a little over 300 bytes.
Both the Date::Manip::Base and Date::Manip::TZ objects are reused
by any number of Date::Manip::Date objects. They can almost be
thought of as global data, except that they are accessible in the
standard OO manner, and you are allowed to modify them on a per-
object basis which WILL mean that you have to store more data. If
you work with multiple configurations (see Date::Manip::Config),
you'll need multiple Base and TZ objects. However, most of the time
you will not need to do this.
The Date::Manip::Date object is a bit larger than suggested in the
complaint, but it should be noted that Date::Manip actually stores
the dates in a number of different formats (a string of the form
YYYYMMDDHH:MN:SS and a list [YYYY,MM,DD,HH,MN,SS] in the time zone
it was parsed in, the local time zone (if different) and GMT. By
caching this information as it is used, it has a huge impact on the
performance.
So, Date::Manip in typical usage consists of one 100K
Date::Manip::TZ object, one 15K Date::Manip::Base objects, and any
number of small 300 byte Date::Manip::Date objects.
Date::Manip::Delta objects are even smaller. Date::Manip::Recur
objects are also small, but they contain any number of Date objects
in them.
I am certainly open to suggestions as to how I can improve the OO
interface... but I don't believe it is a dud. While I'm not an
expert at OO programming, I believe that I followed pretty standard
and accepted procedures for accessing the data.
Please refer to the Date::Manip::Objects document for more
information.
Date::Manip has an inconsistent interface
I've gotten a few complaints that the interface to Date::Manip is
inconsistent... and I agree (at least when using the functional
interfaces).
Date::Manip was originally written in an unplanned way... as a
need/want came up, it was extended. That's not the way to write a
major package of course, but it wasn't expected to be a major
package at the start.
As it became more and more widely used, I too wished for a more
consistent interface, but I did not want to break backward
compatibility for everyone using it.
When 6.xx was written, I spent a good deal of time trying to make a
very standard interface, so I do not believe that this complaint
can be applied to the OO interface (though I'm interested in
suggestions for improving it of course).
As far as the functional interface goes, I'll continue to support
it in a backward compatible (and therefore inconsistent) form. I'd
encourage the use of the OO interface whenever possible.
BUGS AND QUESTIONS
If you find a bug in Date::Manip, please send it directly to me (see
the AUTHOR section below). Alternately, you can submit it on CPAN.
This can be done at the following URL:
http://rt.cpan.org/Public/Dist/Display.html?Name=Date-Manip
Please do not use other means to report bugs (such as Usenet
newsgroups, or forums for a specific OS or Linux distribution) as it is
impossible for me to keep up with all of them.
When filing a bug report, please include the following information:
Date::Manip version
Please include the version of Date::Manip you are using. You can
get this by using the script:
use Date::Manip;
print DateManipVersion(1),"\n";
or
use Date::Manip::Date;
$obj = new Date::Manip::Date;
print $obj->version(1),"\n";
Perl information
Please include the output from "perl -V"
If you have a problem using Date::Manip that perhaps isn't a bug (can't
figure out the syntax, etc.), you're in the right place. Start by
reading the main Date::Manip documentation, and the other documents
that apply to whatever you are trying to do. If this still doesn't
answer your question, mail me directly.
I would ask that you be reasonably familiar with the documentation
BEFORE you choose to do this. Date::Manip is a hobby, and I simply do
not have time to respond to hundreds of questions which are already
answered in this manual.
If you find any problems with the documentation (errors, typos, or
items that are not clear), please send them to me. I welcome any
suggestions that will allow me to improve the documentation.
KNOWN BUGS
None known.
SEE ALSO
Date::Manip - main module documentation
LICENSE
This script is free software; you can redistribute it and/or modify it
under the same terms as Perl itself.
AUTHOR
Sullivan Beck (sbeck@cpan.org)
perl v5.16.3 2014-02-08 Date::Manip::Problems(3)
