hpr-knowledge-base/hpr_transcripts/hpr1286.txt

Episode: 1286
Title: HPR1286: iCalendar Hacking
Source: https://hub.hackerpublicradio.org/ccdn.php?filename=/eps/hpr1286/hpr1286.mp3
Transcribed: 2025-10-17 22:57:00

---

Hello, this is Dave Morris. Today I want to talk to you about a problem I encountered.
Last year it was in fact in 2012 and some of the research I did and experience that I
tried to carry out to fix the problem. I didn't quite get there but hopefully you'll find the
voyage interesting. So the original problem was that back in 2012 Ken Fallon tried to use Google
Calendar to set up an event for recording the monthly community news shows on HPR. He wanted to be
able to send out invitations to people using that means. He wanted to set it up so that the message
went out the event was set for the Saturday before the first Monday of each month. Quite surprisingly
he didn't find a way to do this and ended up deleting the attempt going for another method.
This puzzled me because you'd think that's what calendaring was all about. I looked at the application I
use which is Thunderbird with the lightning plug-in to see if I could do it but I couldn't find a way
either. I'll do surprising. I'm sure Kenneth looked at all manner of ways of achieving this.
So I thought I'd do a bit of research about what might be going on here. I was aware that
calendars like Google Calendar and many more use the I Calendar format to represent events.
So I thought I'd try and find out more about this. I've come across it before and I've always wondered
how this stuff works, how do you represent an event and all the other stuff that you can do
when setting up a calendar. So I grabbed a copy of the RFC that covers this which is 5545
and tried to absorb enough to get a vague idea of how it defines recurrent entries of this sort.
It's pretty heavy going. I must say I haven't read it in my office. It's about 9,000 lines of
pretty dense text but it's possible to grab a copy and I've left a link in the show notes for you.
So I discovered that there are two primary methods of defining recurrent events within
I Calendar. There's a command, I suppose you call it a command, which is R Rule,
spelled R-R-U-L-E and R-Date. The R rule, sorry, property is the more powerful of the two
and of course the more complex. The description of R rule is long involved but in the context of
this problem I could see how to define the first Monday of every month which is if you go and look
at a calendaring application such as Thunderbird it's there in the GUI. You can say I want this
to repeat every month on the first Monday and the rule underneath it is I've written it in the
in the show notes but this is simple enough that I can read it out to you. It's R-R-R-R-R-U-L-E
colon frequency that's F-R-E-Q equals monthly semi colon by day B-Y-D-A-Y equals 1-M-O.
Now fairly obviously the frequency equals monthly simply means that the rule repeats every month
and by day equals 1-M-O then means that the day every month is the first Monday so that's
pretty simple quite neat actually. So most calendar applications are well able to deal with this
sort of specification and seems to be the way in which most recurrent events are defined. You can
dump calendar entries out into devils and go and look at them if you're interested.
So I thought I'd have a try at using this i-calender stuff to do what was needed.
Obviously the first Monday of the month is not what we want to do. We wanted this Saturday before
that but then I found that the i-calender syntax doesn't have an obvious way of subtracting two days
to get this Saturday before. You'd think it'd be easy enough to say first Monday of the month
take away two days and there you go it's not quite as simple as that.
So I discovered that the the by day part of the rule can be followed by a list of days of the
week which are represented as two character strings M-O for Monday, T-U for Tuesday etc.
And you can also proceed each of these days with a digit meaning the first Monday, second Monday etc.
There's a further part to the rule which is by set pause B-Y-S-E-T-P-O-S
S which modifies the by day rule and it's followed by separated list of values, comma separated
list of values which correspond to the in the currents within the set of events. That's pretty
complex to say but we take the rule I had before and modify it a bit. So we've got R rule equals
frequency equals monthly by day equals now here we can we're going to put a list Saturday and that's
S-A comma Sunday S-U comma one M-O so that's a list that says the first Monday of the month
and the list also contains the Saturday and Sunday before it. We follow that with
semicolon B-Y-S-E-T-P-O-S by set pause equals one so the intention here is to say
every month go and find the group of days which contains the first Monday
that group of three days the Saturday and Sunday before it and take the first.
So that's actually quite nice, quite elegant, a little bit convoluted but there you go.
I was quite surprised to find that this actually worked once I'd actually constructed this rule
and plugged it into a calendar but I found that it has a fatal flaw.
If the three days in the by day group are all in the same month it works fine but if either
the Saturday or the Sunday are in the previous month something or other fails to backtrack
enough and drops the event on the wrong day it's as if the algorithm has been written only to deal
with dates within a given month not to span one month to another.
Well even if this worked reliably I suspect that many calendar applications couldn't define it.
The interface certainly with Thunderbird and Lightning the interface does not have the capability
to specify something that's complicated. Anyway in the show notes I've put an example of the
full eye calendar entry that I broke to a file and plugged into Thunderbird to get the effect
that I just described so you can have a look at that and see if you can get it to work with your
calendar. Maybe there's a calendar out there that does actually do the right thing be most
interested if that were the case. So rather than give up entirely at this point I thought it
have a go at another experiment. I discovered that there's an alternative way of specifying
repeated dates though not in such a nice compact way as the R rule. The R date specification I
mentioned before allows you to define a number of events by simply specifying them they don't have
to follow any particular rule and there can be anything. Now there are ways of doing this you can
actually say within calendaring applications one sort of looked at anyway that you want it to
occur on these specific dates though making that happen seems to be rather involved.
Anyway I thought I'd have a go at building a test calendar containing the next 12 community
news events and constructed this by hand you'll be surprised to find that there's a plug-in
for them which recognizes the I calendar syntax so that makes it quite nice and prevents you
making mistakes. So rather than do this I thought I'd put together a pulse script to
generate a block of events using our date and there's a reference to this script in the show notes
you can go and examine it on the show notes page and there'll be a link to a downloadable
copy if you wish to grab it and play around with it. The script uses the pearl modules date calc
and data ical date calc is used to perform date calculations and the other one data ical
generates i calendar data so you can run the script it's called mate calendar and pipes it's
it centers out put to stand it out so you can pipe that to a file there's an example of it being
used in the show notes by default it generates 12 appointments and there's an example of
a set of these appointments in the show notes which you can load into thunderbird and any other i
calendar based calendar I assume and get the desired result you can't edit them in any
certainly not in thunderbird you can't edit the sub-events you can just delete them
so you couldn't move one of them to another day as far as I can make out but the idea works although
it's a hell of a convoluted way to get calendar events i did try this with k organizer i'm a kde
user and that's the only other calendar i had available to me and it sees the multiple dates
as multiple discrete events rather than a single recurring event but there's some hope in this
but it's still not not ideal but maybe better than the the our rule thing that gets the date wrong
so i thought i'd have have a look for and see if i could find another way of doing this so i
went for another experiment which where i looked at other calendaring systems
and there are the two that i looked at do seem to have ways of handling this type of date
specification better for example i've used a tool called pcal for many years which generates
post-crip calendars into which you can add events and then you can hang it on the wall or carry
around with you as i used to do and it's it syntax can quite reliably specify the saturday before
the first monday of each month and you simply type the expressions saturday before first monday in
all which means every month the saturday before the first monday it does does a fine job of
computing it but not particularly useful unless you like carrying around paper calendars
there's another tool that i looked at which is called remind which can do the same thing
and i won't read out the expression in this case have a look at the show notes you can see
the specification to to generate events like this but the interesting thing about remind is
that it comes with the tool which can generate i calendar data it's called rem2 ics
it expects output from the remind command and then it generates i calendar data from it so
i give an example um command pipeline to uh to show how uh you could generate 12 meetings from
the reminder that i've specified there um and uh produce uh an i calendar file from it so i discovered
that this uses our date um as my pulse script does to to generate such a set of events
um and does quite a nice job of it but uh it's it's it's still not the the thing that i wanted
so the conclusion after all this experimentation is that uh the i calendar specification should
be able to generate appointments of the type that we need using the the r rule specification
because you'd want that because you could generate infinite appointments that just character the
rule just keeps operating forever it should be able to do this but the small number of
calendaring applications i've checked it against don't seem to be able to do it for some reason
it's uh the only conclusion i can draw is that they've not been implemented properly
so other tools that don't use i calendar have less difficulty representing such things but uh
they're not quite the same thing they're not going to call you to to a meeting and all that sort
of good stuff that uh that i calendar thinks are supposed to do
so i'd be interested if anybody has any other experiences around this area um some thoughts about
how this could be solved more effectively and just let me know so it's a bunch of links in the
show notes that hopefully will give you some help if you want to follow this up any further
but otherwise that's it okay bye
you have been listening to Hacker Public Radio at Hacker Public Radio does our
we are a community podcast network that releases shows every weekday on day through friday
today's show like all our shows was contributed by a hbr listener like yourself
if you ever consider recording a podcast then visit our website to find out how easy it really is
Hacker Public Radio was founded by the digital dog pound and new phonomicum computer cloud
hbr is funded by the binary revolution at binref.com all binref projects are proudly sponsored by
linear pages from shared hosting to custom private clouds go to lunar pages.com for all your hosting
needs. Unless otherwise stasis today's show is released under a creative comments
attribution share a life lead us our lives.