Dynamic Shift

[EricDallAgnol]

Hi team 👋

Currently, when it comes to implement shift, they are two functions exposed :

• tt.function.shift
• tt.function.date_shift (for LOCAL_DATE)

However, the offset provided is always here a 'static' value.
Is there a way to handle that dynamically ?

I'm doing a POC and I would like, through a parameter hierarchy to be able to shift on the fly in my measure

Let's say I have a VaR 99% measure, my current date and the date with which I want to compare my VaR (this second date value comes from a parameter hierarchy in my example) :

In this example I have a shift of -1 day thus I would like my VaR 99% Shifted measure to be shifted by one.
I'm wondering if there is a way to provide the offset based on my date_diff measure, so that if I select the date to be compared to another and my date_diff becomes, let's say -5 this time, then I would like my VaR 99% Shifted measure to be shifted by -5 (and to have only ONE VaR 99% shifted measure, not one shifted by -1 and another by -5)

Many thanks,

Eric

[tibdex]

Hi Eric,

Indeed, passing measures to date_shift() or shift()'s offset parameter is not supported. As a workaround, if you have a small set of known in advance different offsets (let's say around 10), you could do something like that:

tt.switch(
  m["date_diff"],
  {
    -1: tt.date_shift(m["VaR"], h["RiskDate"], offset="P-1D"),
    0: m["VaR"],
    ...,
    5: tt.date_shift(m["VaR"], h["RiskDate"], offset="P5D")
  }
)

You could even create the switch() cases with a dict comprehension to have less things to type.

[EricDallAgnol]

Hi Thibault,

Thanks for your answer. Indeed the switch() can do the job - in my case it should be acceptable to handle that within predefined ranges.

Out of curiosity, would passing measures to date_shift() or shift() is on the medium/long term roadmap ? It would be a nice way to use the date_diff() function in my opinion.

Thanks !

Eric