The Ellipsis/Not Implemented Error In M

Recently I was creating a parameter in Power BI Desktop and had it configured something like this:


I didn’t bother to choose anything in the Default Value dropdown box, and when I looked at the code for the parameter in the Advanced Editor I saw this:

 List={"a", "b", "c"}, 

I was interested to know what the ellipsis symbol (three dots …) in the DefaultValue field in the record meant, and looking in the Language Reference I found that in M it can be used for two purposes: as an open record marker (maybe something for a future blog post) and, as in this case, as a quick way of returning an error. The language reference says that it is directly equivalent to the following expression, which returns a “Not Implemented” error:

error Error.Record("Expression.Error", "Not Implemented")

But from what I can see, it actually returns a “Value was not specified” error instead. Here’s another example of how it can be used in a function definition, and what it returns:

    MyFunction = (x,y) => if x>y then true else ...,
    Output = MyFunction(0,1)


It’s not something I think I’ll be using in my own code, but it’s good to know what it means!

Creating Sequences Of Integers And Characters In Power BI/Power Query Lists

When you’re writing your own M code you often find yourself needing to create a list containing a sequence of numbers or characters. Luckily the M language allows you to do this very easily when you are defining lists by using expressions in the format


For example, imagine you want to create a list with all of the integers between 1 and 5. Instead of writing

{1, 2, 3, 4, 5}

You can write the following:


and it will return the same list:


You can also use this format in more complex list definitions, for example

{1..3, 5, 7..9}

Returns the list

{1, 2, 3, 5, 7, 8, 9}


When you’re using this technique you must always put the lowest integer first and the highest integer last; if you don’t do this you get an empty list. So, for example, the expression


Returns an empty list:


It’s also possible to use this technique to create lists of characters. For example, the expression:


Returns a list containing all of the lowercase letters of the alphabet:


The first character in the expression has to have the lowest Unicode value and the second character has to have the highest Unicode value, and the sequence of characters returned is the list of all characters with Unicode values in that range. As a result, the expression


Returns the list

{"#", "$", "%"}


And the expression


Returns an empty list because the Unicode value of “a” is greater than the Unicode value of “Z”.

This technique doesn’t work for decimal numbers, dates or other data types. If you want a sequence of values of these types you need to use functions list List.Dates() and List.Numbers().

Lists are, of course, used all over the place in M. Building on my recent post on using #table() to create tables with no data source, here’s one last example of using lists containing sequences to create a simple table with three columns and three rows:

#table({"A".."C"}, {{1..3}, {7..9}, {11..13}})


Two Upcoming Power BI Webinars

This is just a quick post to let you know about two webinars I’m presenting soon. First, on the 21st of June (today!) at 1pm PST I’m presenting a webinar on “Introduction to M” as part of the Power BI community webinar series; more details on it and how to register can be found here:

Second, I’m presenting a webinar with Pyramid Analytics about their on-premises BI solution and how it integrates with Power BI on the 30th of June called “The Public Cloud Is Not For Everyone”:

[Full disclosure – I’m being paid for this – but it won’t be marketing fluff, just honest discussion, similar to the webinars I did with Pyramid last year]

MDX, DAX, Power BI, Cortana Intelligence And SQL Server HA Training Courses

If you’re looking for expert-led training courses in Microsoft BI and SQL Server, check out the courses we have coming up in London at Technitrain:

If you’d like to keep up-to-date with all our new course announcements, please join our mailing list!

Power BI Diagnostics, Trace Logs And Query Execution Times (Again)

I’ve blogged a few times now about how to monitor Power BI query execution times, most recently here, and while using functions like DateTime.FixedLocalNow() is all very well I’ve never been happy with the fact that you have to alter your queries in order to be able to time them. Using the Power BI trace logs always seemed a much better option – and since the log files are in a reasonably easy to understand text format, you can use a Power BI query to load data from them. I wrote a post on importing data from the trace logs (using Power Query) here back in 2014 and Rui Romano has a great post on doing the same thing for Power BI here. The big problem with the Power BI trace logs, though, is that there is too much information in them: they’re really meant for Microsoft internal use, I guess, and not for the likes of you or me. In fact it’s all low-level data in there and there’s no way to see (as far as I know) the overall time taken for a query to run.

However, last month when I was looking for new M functions using the old #shared trick I came across the Diagnostics.Trace() function which allows you to write to the Power BI trace log direct from your query.  After a bit of experimentation, I found out that this function is the key to a truly elegant way of logging query execution times. Here’s a proof-of-concept that I’ve put together to show how it could be used…

First of all, consider the following query called LongQuery:

    Source = Function.InvokeAfter(
                     #duration(0,0,0,1 + Number.RandomBetween(0,5)))


It returns the text “Hello” and uses the Function.InvokeAfter() and Number.RandomBetween() functions to wait anything from one to six seconds before doing so. Let’s pretend that this is a real query that does something complex and that we’re trying to tune it.

Rather than alter this query, instead the best thing to do is to create a new query that calls this query through Diagnostics.Trace(). Here’s the code:

    Executing = Diagnostics.Trace(
		#"Trace Message", 


I’ve called this query Diagnostics. All it does is return the same value as LongQuery but in doing so it also writes a single message to the trace log. #”Trace Message” is a text parameter that allows you to specify what message to write – you might want to put some kind of version number in here, or a description of any changes you’ve made to your query while tuning it:


You should also make sure that the Diagnostics query doesn’t get loaded into the data model by making sure the Enable Load option is deselected:


One important thing to remember is that for the Diagnostics query to actually log anything, you need to have the Enable Tracing option turned on in the main Options dialog (you can find this on the File menu in the main Power BI window under Options and settings/Options):



You will also need to remember to turn this on every time you open Power BI Desktop.

With that done, you just need to hit the Refresh Preview button to refresh the Diagnostics query in the Query Editor window:


The Diagnostics query will run, calling the LongQuery query and making an entry in the Power BI trace log. You can find the directory with the trace logs in them by clicking on the Open Traces Folder link in the Options dialog shown above, and you can open the any of the trace log files using your favourite text editor.


Loading this trace file into Power BI is pretty straightforward using code similar to that in Rui’s or my blog post. The difference is now, instead of loading everything, you can filter the rows to only load the events created by Diagnostics.Trace(). as far as I can see you can identify these rows because they have the value


Somewhere inside them, as you can see from the last screenshot. Most importantly of all, the duration logged for this event is the amount of time taken to execute the query.

Therefore, with some Power BI M magic, you can load all the data from all of the log files in the log directory, filter the rows to only those created by Diagnostics.Trace(), and end up with a query that gives you something like this:


The Execution Index and Date Time columns tell you when the query was executed, the Message column contains the value of the Trace Message parameter at the time the query was executed, and the Duration column tells you how long the query took to execute. You can then, of course, load this table into the data model and build a report that helps you analyse these execution times:


To sum up, the workflow for tuning your query is:

  • Make some changes to the LongQuery query that hopefully make it faster
  • Update the Trace Message parameter with some notes about which version of the LongQuery query it is that you’ll be testing
  • Click the Refresh Preview button for the Diagnostics query to test how long LongQuery now runs for
  • Refresh, or load, the query that reads the data from the trace logs so you can see how all of your changes have affected query execution times

Some other closing comments:

  • As I said before, this is just a proof-of-concept and I although haven’t tested it thoroughly but it seems to work.
  • Microsoft may well change the format of its trace logs or make other changes in the future that could break this code shown here.
  • Although I haven’t tried it, this approach should work for Power Query too with some minor modifications.
  • Since it relies on you turning on tracing in Power BI Desktop, this won’t work after you have published your reports. In fact I’ve deliberately structured the queries here so that logging only takes place when you click the Refresh Preview button for the Diagnostics query in the Query Editor window – I found this was the best way to reduce noise in the log files, especially given the fact that a query can be executed multiple times when data is actually being loaded into the data model.
  • That said, the Power BI Enterprise Gateway has its own log files, so maybe you could adapt this approach somehow to work with them?
  • I suspect there are lots of other uses that Diagnostics.Trace() could be put to for even lower-level debugging and analysis of queries, but that’s something for a future blog post.

You can download the sample code as a Power BI template from here. When you open the template you’ll be prompted to enter two parameter values: the message to record for your events in the trace log and the path of your trace log directory.

Creating Tables In Power BI/Power Query M Code Using #table()

After my post earlier this week on creating current day/week/month/year reports in Power BI a few people asked me for a more detailed explanation of the way I was creating tables without using a data source in my M code. This is something I find myself doing quite a lot when I’m loading data with Power BI and Power Query, and while there are several ways of doing this I find that using the #table() intrinsic function is the most elegant option.

Let’s look at some examples. The following query returns a table with two columns (called “First Column” and “Second Column”) and two rows containing the values from 1 to 4:

#table({"First Column", "Second Column"}, {{1,2},{3,4}})


No data source is needed – this is a way of defining a table value in pure M code. The first parameter of the function takes a list of column names as text values; the second parameter is a list of lists, where each list in the list contains the values on each row in the table.

In the last example the columns in the table were of the data type Any (the ABC123 icon in each column header tells you this), which means that they can contain values of any data type including numbers, text, dates or even other tables. Here’s an example of this:

 {"First Column", "Second Column"}, 


While this is flexible it’s not exactly practical: in almost all cases the Any data type is a bad choice for loading data, and you need to explicitly set the data type for each column. You can set data types for columns quite easily as a separate step, but it is also possible to set column data types using #table():

 type table
        #"Number Column"=number, 
        #"Text Column"=text,
        #"Date Column"=date


In this example the first parameter is no longer a list of column names but a declaration of a table type that not only has column names in but also column types. You can see from the icons in the column headers in the screenshot above that the column called “Number Column” has a data type of number, “Text Column” has a data type of text, and “Date Column” has a data type of date.

Of course if you need a fixed table value in Power BI you could use the “Enter Data” button or, if you’re using Excel and Power Query you could create an Excel table and then use the Excel.CurrentWorkbook() function to load the contents of it; if you or your end users need to edit the values in your table easily then you should use one of these two options. On the other hand if you don’t want users to be able to edit the values in the table or, more likely, you are generating the contents of your table using functions that return lists (as in my previous post) then #table() is the way to go.

%d bloggers like this: