An interesting bit of date relevance

Content Authoring
1

Was working with a colleague today and he asked a great question. Why do some of these not exist, and is this a bug?

Q: month of (current date - 0 * month) as string
A: March
Q: month of (current date - 1 * month) as string
E: Singular expression refers to nonexistent object.
Q: month of (current date - 2 * month) as string
A: January
Q: month of (current date - 3 * month) as string
A: December
Q: month of (current date - 4 * month) as string
E: Singular expression refers to nonexistent object.
Q: month of (current date - 5 * month) as string

It turns out that when you take March 31, 2025 (a date) and subtract a month… well… there is no February 31 (another date) so the month of the non-existent date also does not exist.

Fortunately, there was an easy solve this. Simply do the Month Math vs the Month of the date instead of the day.

q: current month - month * 0
A: March
q: current month - month * 1
A: February
q: current month - month * 2
A: January
q: current month - month * 3
A: December
q: current month - month * 4
A: November
q: current month - month * 5
A: October
2 Likes
2

Note that current date will not be accurate if you leave your Fixlet Debugger running more than 1 day:

Q: current date
A: Fri, 14 Mar 2025
T: 0.069 ms
I: singular date

Q: now
A: Mon, 31 Mar 2025 13:41:09 -0600
T: 0.041 ms
I: singular time
1 Like
3

Update on the relevance. my colleague came back with a twist, because he wanted to get the Year of 4 months ago and was hitting existence again.

Solved with

q: year of (current month_and_year - (4 * month))
A: 2024

Generally, you can add months to “month like objects” where you do not have to know “how many days in a month”. As soon as you get a date like object, to add a month, you have to know how many days to add. You can add a specific month to a date, but not a general month object.

@itsmpro92 good times indeed! I have only hit that once in my years of BigFixing with a student doing a 3 day class. It was very puzzling when we came across it, but super rare, so it has never bubbled to the top of the urgent pile.

4

You’ve had your Fixlet Debugger open for 17 days?

:exploding_head:

5

It’s my lab environment :man_shrugging:. @brolly33’s comment reminded me that I had run into this issue while doing some content development around current date.

6

Not a bug! The language and date math are correctly erroring that no “2025-02-31” exists.

In the world of calendar interoperability, there is no “right” way to do this. In implementation, however…

  • Some systems appear to choke, as relevance does here.

    • This is generally regarded as bad.

  • Some systems error on “the prior month has 28 days and you’re on the 31st so we’ll punt and give you the last day of the prior month”: 2025-02-28.

    • This is more common.
    • Extra fun on leap years!

  • Some systems error on “the prior month is 3 shorter than this one, so we’ll punt and give you that month plus 3 days.” : 2025-03-03

    • This is less common.
    • The school of thought here emphasizes that “days” and “months” are different things. Best to give you all the days.
    • Extra super fun on leap years!

Fundamentally, however, it’s an implementation choice. The language should do exactly what it’s doing: report an error. The coder needs decide what you probably “meant”, know date math and date edge cases, and decide how to handle them.

As always, I point to this very amusing site.
https://yourcalendricalfallacyis.com

2 Likes
7

Regarding the Debugger issue, I found recently that current date may not be what you had in mind, but now as local date probably is.

2 Likes