These dates were derived via Groupby Mean of how ever many properties closed each month of each year. Obviously, if nothing closed that month, then no mean date derived.
I’m not worried about extreme accuracy for splitting the distance of exact days or anything. If I had some system to state the 15th of each missing month (and year), that would be enough to fill the gaps for a time series presentation of the data.
If you are happy to use linear interpolation to fill in the missing dates, then it would seem to me that turning dates into a serial number representation and then using Missing Values is a possible way to achieve this.
I have a component on the hub that I have just modified (as before it only accepted Local DateTime columns, but it will now accept a Local Date column too)
This can produce a “unix serial number” for the date, based on seconds since 1970-01-01. With this as a serial number, the Missing Values can then perform linear interpolation on it. Turning that into a Long , the standard node Unix Timestamp to Date&Time can convert the number back to give you interpolated dates:
To do this without the component, calculate the number of seconds since 1 Jan 1970 and multiply the result by -1, or take the abs() value using Math Formula, as follows:
Hi @takbb I ran your WF linked above (thanks for that; it’s really nice to have the settings to work with). But my end dates ended up substantially in the future, like years 2080, 2083, etc. I will double check to see if I missed something on the settings
