x_PricingBlockSize is the column name in FOCUS dataset, but inside the reference data CSV file, it is the column name is PricingBlockSize.

Given the mapping doesn't happen in this column, what do you think about only calling out the division here and leaving the mapping for the x_PricingBlockSize column?

Same as before, please use PricingBlockSize.

Same comment as before...

Clarify that the OriginalValue column in the reference data file is what is used to map from ResourceLocation.

Good catch! Do you think this is clear enough? 🤔

I believe for the Services reference data file, we need to map based on ConsumedService and ResourceType, based on this documentation - https://microsoft.github.io/finops-toolkit/data#%EF%B8%8F-services. It's not unique just by ResourceType.

Yes and no. I need to go look at the updated mapping we use internally, but there are only 1 or 2 services that need a ConsumedService mapping. The rest are all ResourceType. I have a task on my backlog to merge Services and ResourceTypes and just call out the exceptions separately. Any thoughts on how you'd prefer to see that here?

Same here.

See previous comment

Here, it mentions focus:RegionName. Should the value also be looked up from the Regions reference data file via the ResourceLocation value and mapped to RegionId?

We could. The reason I used the RegionName reference is because doing another lookup takes more time. This is just faster and aligns with the general guidance we have from the ARM team about how to handle location IDs.

How does one obtain focus:RegionName? It's not a column. Confused on that.

I believe this can be obtained from the ResourceId, but let me know if I am wrong:

regexp_extract(ResourceId, '.*/([^/]+)$', 1)

Take the last segment of the / delimited value.

It can, but it's not that simple. Nested resources technically have a name that includes all levels in the resource hierarchy. So a SQL db (under a server) would be "MyServer/MyDatabase".

I forgot that this isn't in the MCA dataset today 🤔 Doing the "right" thing is a bit onerous and probably wouldn't be clear here. If I remember correctly from doing this in KQL, it requires a few lines of code. Your approach is simpler, but I hate to rely on regular expressions. What do you think about this:

Not sure if the [-1] reference will be clear to everyone since I'm only aware of 2 platforms that use that 🤔 Maybe the text is good enough? (Although that should probably be in a footnote and not in the "transform" column.)