I don't see functions showing up in the service either. It looks like the UDF icon isn't an official one yet. In Desktop the fx part is a different font face than the fx for Expressions and I don't see a version for it here:
github.com/microsoft/fl...
The written documentation is great. The piece I was missing was updating my Compatibility Level.
docs.tabulareditor.com/te3/tutorial...
Screenshot of a video from the Tabular Editor 3 documentation titled "Create DAX Functions in Tabular Editor"
I'm was trying to figure out how to create DAX UDFs in #TabularEditor. Imagine my surprise when I found a Kurt Buhler video claiming "This video teaches you how to add DAX UDFs in Tabular Editor 3" but doesn't create a single function during the entire 9 minutes. 🤯
@marcorus.bsky.social
This definitely resonates with me. AI is fantastic for overcoming technical barriers and quickly building proofs-of-concept, but productionalizing them takes real experience.
It doesn't need to all fit into context simultaneously. The trick is accessing the necessary pieces of context as they are needed. So far, humans are much better at this--likely by necessity since our short-term working memory is so small. Computers needn't be as restricted.
This shows how much potential there is for AI-run companies to outcompete humans--though likely not for a few years. The possible improvements in global understanding, coordination, and low-latency, high-bandwidth communication are enormous.
@jaypowerbi.bsky.social @itsnotaboutthecell.com Please tell me this is just a new person taking a shortcut, not standard operating procedure.
Microsoft is using AI slop to power their Community Support Team
Example:
community.fabric.microsoft.com/t5/DAX-Comma...
I read it a week ago and asked in the comments what LLMs Fabric Copilot is based on but it must have gotten caught in a spam filter.
I'm confused. Union Pacific's Dallas Yard, the Cadiz St. & S Lamar St. intersection (~1.5 miles from 32.761 N, -96.817 W), and the pin at the end are all separate locations. Which is correct?
Pretty sure this is what's going on. It can't process the whole thing with the same focus as a specific zone.
They've been disappointing so far but it's not over. They don't need to build frontier models, just figure out how to integrate one with their ecosystem.
I've often wanted a good way to search/scrape large volumes of MS Community forums to do classification of common questions but haven't found good solutions.
Is there something you can do for community forums?
Apparently, COBOL doesn't even have a date type but a 6-digit date representation does have a plausible rationale for a Social Security 1875 epoch date, so there may be a grain of truth that's been garbled.
retrocomputing.stackexchange.com/questions/31...
A decreasing log x-axis? If you want cheaper to the right, then use tokens/$ instead.
Charles Bridge, Prague
It's both. I don't like holding details in my head. I'd rather see them laid out plainly on a screen.
There are custom visuals that make it super easy without any extra modeling or DAX. I hope Microsoft brings that functionality into native visuals eventually.
Hitting those milestones feels great! What app are you using?
Too bad it's not as easy as acquiring 3rd party tools like Deneb or Inforiver. Those teams have been pretty impressive.
Personally, most of the "top items" don't make a significant impact to my own work. However, when I see several top ideas from 2015 that are "in our upcoming roadmap" since 2017 it looks like the site is defunct. Updating the status of top ideas at least once per year would improve this look.
I think the fundamental issue is that users don't feel like MS reliably responds to their needs. MS is doing Copilot and Fabric stuff that nobody asked for instead of addressing basic things that have been top 20 on the ideas site for 5-10 years. That site is a bad joke where feedback goes to die.
#8 (Roche's Maxim) is the golden rule
Yep. Not needed but I like it keeping these filters more compact and matching how Power BI auto-generates its queries that populate visuals.
I think this may have changed since 2017. I tested the file from the article you linked and the TREATAS version has simpler physical and logical query plans as well as slightly better performance.
See also the comparison here: www.sqlbi.com/articles/pro...
Rewrite a FILTER as TREATAS (for use in SUMMARIZECOLUMNS queries):
Before: FILTER ( ALL ( Table1[Col] ), Table1[Col] IN { "X", "Y" } )
After: TREATAS ( { "X", "Y" }, Table1[Col] )
Simplify compound filters on the same column
Before: Product[Color] = "Red" || Product[Color] = "Blue"
After: Product[Color] IN {"Red", "Blue"}
@emollick.bsky.social
Any hypotheses about why ChatGPT throws errors on certain names?
www.reddit.com/r/ChatGPT/co...
Thanks. I do have the paid version but I've heard cool thing about Semantic Link from multiple people.