I ran across this bug feature while processing some NOAA data that I have been playing with off and on for a few years. If your text files have a single quotation mark in them, i.e. “, the Query Editor in Power BI will not handle it well.
Lets see what happens.
First, I loaded a text file into Power BI that has a list of country names in it. In this case, it is the country names that will appear in the weather station data that I am playing with. The file is a simple fixed-width file with 291 entries that looks like this:
If I import the entire file, I end up with 291 rows in the Query Editor that looks as expected.
Now, if I add a simple ” to the text file as shown below, save it and refresh the query editor, things go sideways quickly.
Here is what it looks like in the Query Editor:
Once I load the file, it looks like this:
If I add two quotation marks, it works as expected:
But if I add three, all bets are off again.
As a final test, lets add two rogue quotation marks to our data like this:
That gives me these two:
I don’t have access to source code anymore but it sure seems that in the code for the query editor, as it is processing text data, once it ‘sees’ a quotation mark, it ignores all carriage return/end of line statements until it ‘sees’ another closing quotation mark.
Now this may not seem like a big deal and in the examples I showed, its pretty easy to see that something is amiss, but in my data files this took a while to find. I was processing hundreds of thousands of rows of fixed width text data so I didn’t see the behavior in the query editor. Once the data was loaded, the stray ” that I had in my data occurred deep in the set of data and since it doesn’t throw an error, this really screwed up how the data was being loaded after the single quotation mark was parsed.
By the way, I tested this in Excel and I saw the same behavior. No surprise I guess.
So is this a surprise to anyone? Is this something that everyone that has ever imported text data into Excel or Power BI already knew about? I get it that it’s an edge scenario that you find a single quotation mark in English fixed width text but this showed up in raw data files from a US government agency so it does happen.
Tell me in the comments – should I have already known this?
I am not sure why but I am suddenly getting a lot more unsolicited calls and emails from recruiters. A good chunk of them are straight spam. Anyone who would send me job postings for SharePoint (Its been four years since I supported it) or C# developer (12+ years since I did it for a living) obviously hasn’t spent more than 5 seconds with my actual resume.
I am always intrigued to see the Power BI ones that come through because for the most part, they still seem to be stock Microsoft Business Intelligence job descriptions from five years ago with the words ‘Power BI’ hastily scribbled at the bottom.
Power BI is a significantly different animal than the standard MS BI solutions that have been out there for decades now and it takes a special skill set to create really effective and powerful BI solutions. The skill set and knowledge base are unique to this product so I thought I would take a stab at the top 5 skills that a generic Power BI developer needs to know to be effective.
In a previous blog post, I have talked about the various diagnostic data that Power BI Desktop has available for troubleshooting issues. One of the interesting files generated is ASPerfCounters.JSON.
I originally thought the name came from ASP.NET performance counters but I am guessing that it actually has its roots with Analysis Services.
I was curious to see what data was contained in this file so I used Power BI Desktop to open and parse the file. Power BI Desktop has a JSON file connector so it was relatively simple to crack the file open and see what was inside. I was pretty surprised to see that the file actually had PERFMON data from the time that the file was generated to an hour earlier. The data is collected every 60 secs for the following performance counters:
To load the data into Power BI Desktop, I did the following.
Use the process in the earlier blog post to generate a Frown Snapshot File. Open the zip file generated and copy out the the ASPerfcounters.json file that has been generated.
Use the JSON File connector to ‘Get Data’ and open the ASPerfCounters.json file.
3. The Power Query Editor will open and it will show a list of records. *Note – the following steps are how I imported and cleaned the file, no doubt there are other ways that it can be done as well.*
4. First we need to convert the list to table. If you right click the list column, you will see an option called ‘To Table’.
Select ‘To Table’ and you will get a confirmation dialogue asking for delimiters and how to handle extra columns.
I left everything as default as shown and selected ‘Ok’.
5. At this point, your list has been converted to a table. I selected the expand icon and selected ‘ok’ to the dialogue that appeared:
6. Once expanded, Column1.p is yet another list that needs to be expanded. When you select the expand icon, you get the option to expand to new rows. Select that and will now see that Column1.p contains records that need to be expanded. Expand Column1.p once more time and select both the key and value. You will most likely want to ensure that ‘Use original column name as prefix’ is unchecked.
7. At this point, you have all of your data expanded and accessible but there are some more things to be done to get at least a basic set of useful data.
First, delete columns ‘Column1.i’, ‘Column1.n’ and ‘column1.u’. Nothing useful there.
Next you will want to pivot the key column. Select the key column and choose ‘Pivot Column’ from the Transform section of the ribbon as shown:
When the dialogue pops up, select the default which should have the ‘Value’ column chosen as the ‘Values Column’.
This will display the data in a nice familiar format, date in the first column and columns for each performance counter. Of course, the data looks pretty odd:
The date is displayed in what is apparently a native JSON date format. I had to go hunting on the Power BI Forum to find out how to handle this format. Luckily, Greg Deckler had this one in the bag. The date is stored as milliseconds from 1/1/1970. He lists a way to convert it using a calculated column but I wanted to convert it in power query so I used the following steps.
I selected the date column, then chose ‘Extract’ from the Transform section of the ribbon. I chose ‘Text Between Delimiters’ and used the open and close parentheses as delimiters as shown.
Now I was dealing with a column that was in milliseconds so I set about parsing out the date and time. I first renamed ‘Column1.t’ to ‘DateMill’, converted that column and all of the perf counter columns to actual numbers and then to get the current date, I added a custom column with the following code:
To get the time value in GMT, I created the following custom column:
= Table.AddColumn(#”Added Custom”, “Custom.1”, each Time.From(([DateMill]/86400000) – Number.IntegerDivide([DateMill],86400000)))
At this point, you can spend more time cleaning up the data and charting it as you wish:
So, why is this useful? Well, its cool that you can use Power BI Desktop to parse JSON files. This file might prove useful if you have a large Power BI model that takes hours to refresh. In case of a crash, you can look at this file and see what was going on with your desktop performance. You could also use these metrics to review performance during a refresh in case Perfmon is not your cup of tea.
If you have your model open in Power BI Desktop, you can get a quick dump of all of your Power Query code by doing the following:
From the top ribbon, Choose ‘Help, then ‘About’
2. This will launch a popup window. Choose ‘Copy diagnostics to clipboard’.
3. Open Notepad (or Word, or OneNote, etc.) and paste. All of your PowerQuery M code will be included in the information that is copied.
4. Note also that when you do this, you generate a ‘Frown Snapshot’ file in your diagnostic folder. By default, this snaphshot will have some ASPerfCounter data and ‘FlightRecorderCurrent.trc’, which is your Power BI ‘black box’ recorder.
Performance always is an issue, isn’t it? Throughout a career that wandered through IIS (My web pages are slow), SharePoint (My web parts/lists/search queries/indexes are slow), SQL Reporting (My SSRS reports/stored procedures/excel pages are slow) and now Power BI (My visuals are taking forever to load), troubleshooting slow performance is always a big part of what I do.
Troubleshooting Power BI visuals can be a little tricky. There aren’t any obvious dials or gauges to look at, you can’t spin up perfmon and attach it to Power BI desktop and the logs, while impressive looking, won’t help you narrow in on the poorly written measure that is killing your performance. What I am going to layout next is a quick approach that you can take to not only get a good look at the performance of a report page but how also you can narrow in on the measures that are dragging you down.
Quick Setup Note – I am using the customer profitability sample from Microsoft for my PBIX file. Its visuals load super quickly but its a quick and easy download here.
First things first, your reports and data model need to be in the same PBIX. We will be using Dax Studio to connect to the data model and run a trace so everything we are testing needs to be in the same PBIX. If you have your visuals and data model in separate PBIX files, you will need to recreate your visuals in the PBIX where your data model live.
Create a blank report page. Power BI desktop will load visuals on the report page that it opens when you open the PBIX so in order to capture a true idea of the page performance, you need to create a blank report page and save the PBIX with that page active.
Close and reopen your PBIX file. If you did step one, you should be looking at a blank report page.
Open Dax Studio and the ‘Connect’ screen should open. Select your open PBIX dpcument as shown and select ‘connect‘.
Once connected, click on the ‘All Queries’ button in the ribbon. This actually starts a trace on your SSAS instance that is running in Power BI desktop. Once the trace is ready, you will see ‘Query Trace Started’ in the output window.
Return to your PBIX that you have open in Power BI desktop. Click on the report tab that you wish to trace and let the page fully load. Once the page loads, you can stop the trace by returning to DAX Studio, choosing the ‘All Queries’ tab and selecting the stop button.
Once the trace is stopped, click on the duration column header to sort the queries by duration. As I mentioned earlier, this demo is super fast so the ‘slowest’ query took 21ms but hopefully you get the point. You know have a list of queries that were performed to build your page, along with the time it took to execute each of the queries.
Continuing on, double click on the query text in the ‘Query’ column. The actual code used will show up in the editor section above the output section. Now you can analyze the DAX being called as well as run an individual trace to dig in deeper.
At this point, you can run all of the DAX in the editor or you can highlight and run just sections of it, just like normal in DAX Studio. If you enable the Query Plan and Server timings options, you can capture a trace and see the actual queries that are being passed to the formula and storage engine for processing. Enabling the query plan option does just what it says, it gives you the query plans, both physical and logical, that were chosen to run the queries.
I have a long animated GIF below that shows turning on the query plan and server timings options, setting the ‘Run’ option to flush the cache each time I run a query, then running the query. I then show where you can find the query plan and server timings information. Since I got the whole screen in the GIF, its a pretty lousy resolution but perhaps if you open it in another tab, you can see enough detail.
Questions, comments, suggestions on digging deeper into Power BI visuals/reports performance? Throw me a comment or hit me up on Twitter – @szBigDan.
Gilbert Quevauvillie from FourMoo posted a great blog post identifying all of the different processes that are running when you have Power BI running.
He identifies four different processes that you will see running and a quick blurb on each. I wanted to post a quick follow up to show how the diagnostic logging in Power BI captures output from these processes.
Power BI Desktop has diagnostic logs? Absolutely!
To turn on the logging, enable it from your options as follows:
There is a link that you can click on that will take you directly to the traces folder in case you aren’t the type to memorize logging locations for software.
When you open this folder, \%username%\AppData\Local\Microsoft\Power BI Desktop\Traces, there are actually two sets of logs. The first set is stored directly in this folder and they will only be created once you turn on tracing.
If you open the Performance folder that is in this directory, you will see that Power BI Desktop actually is ALWAYS logging. I am fairly certain that these are a rolling set of logs that are captured in case of an unexpected failure. When Power BI crashes, it will create a FrownSnapShot zip file and a PerformanceTraces.zip file. The latter is a zip of the Performance folder. The FrownSnapShot zip file is pretty cool because it contains the latest SQL Flight Recorder data that has been captured. Interestingly enough, the FrownSnapshot zip also includes a file called ASPPerfCounters.json which looks like Perfmon data that is dumped into a JSON file. Interesting……
The log files that you will find in the Traces folder are as follows.
This file logs information for the Power BI Desktop Application.
This log appears once you go into the Query Editor or refresh your data. This is where Power Query actions are logged during the processing and cleaning of your data.
Once you start a data refresh in Power BI Desktop, you will see multiple log files that start with this name. I am going to assume that Power BI Desktop will spin up multiple Power Query engines as necessary and each will have its own log file.
From my experience, you will only see this log if an exception is captured by the SSAS engine in Power BI Desktop. For instance, I saw this log appear when I was troubleshooting a crash that was occurring during data refresh. This log is also included in the FrownSnapshot zip that is created during a crash.
If you have questions about Power BI Logging or want to add corrections, please leave me a comment!
Quick hitter on this one. I pushed some reports to the Power BI Service lately and noticed that suddenly I didn’t have the ability to drill down into my matrix visualizations. I could right click a row in the visualization and get the option to drill down but I am used to being able to hover over and see the following arrows pop up:
It’s a quick easy fix to get these back. Once you have the March 2018 release of Power BI Desktop, you have the ability to turn off the visual header. This means that while the report is in reading mode, the headers won’t appear when you mouse over the visualization.
Sure enough, when I went into the Reports Section of the workspace and checked the settings, I saw this:
Not sure how that got toggled on for my reports but since I had built a bunch of them from the same template, it’s easy to see how they all that setting applied. Toggle that switch and its back to normal.
I started to really get obsessed with data back when I was supporting SharePoint. SharePoint used to (maybe still does?) have this huge diagnostics database that logged amazing amounts of data by default. I am sure the product team had imagined writing all sorts of cool stuff around it but that feature must have been cut because the database appeared in SharePoint almost as an afterthought. I spent many hours trying to use all sorts of tools to pry knowledge from that database but I never really got too far.
My other data obsession that started around the same time was digging through historical weather data. Ever since seeing this episode of the Simpsons, I wondered what data was being collected by these mysterious weather stations. Years later I finally found my answer when I stumbled across the National Centers for Environmental Information website, found here: https://www.ncdc.noaa.gov/
After some poking around, I found an FTP site where the mother lode of weather information was stored and I have poked and prodded around with it ever since. In part one of this series, I will cover how I automated pulling these files into an Azure Data Lake using Azure Data Factory. As the series progresses, I will cover ETL/ELT and building out of a model for Power BI as well as building some visualizations.
Fits and Starts
In the past, I had used an SSIS package with a FTP task that downloaded all of the files to my desktop. I used a For-Each container to spawn an instance of 7-zip to unzip each of the files and then used an Azure Blob task to upload the files into Azure. This was an effective approach but it hardly seemed enterprise worthy so I started looking into using Azure Data Factory (ADF) to do the same thing.
The obvious approach with ADF would be to create a FTP data set and use a Copy Data activity to pull the files from the FTP server and move them to my desired location. I hit an immediate brick wall however. The NOAA FTP site wasn’t anonymous, but to log in, you use the user name ‘anonymous’ and pass your email address as the password. The FTP data set in ADF will only allow you to choose anonymous or Basic authentication and the FTP server did not like the fact that the ADF FTP data set automatically encrypted the password. Honestly, I got stuck here for a while.
What finally worked
Eventually I found that NOAA had the same data stored on a website that allowed anonymous access. It was off to the races at this point, here is what I put together in ADF.
The website, https://www1.ncdc.noaa.gov/pub/data/ghcn/daily/by_year/, looks like this:
Each file consists of a years worth of data in csv format that has been GZipped. The URL of each file was just the URL listed above + the name of the file.
I wanted to be able to run a copy data action against each file so I brute force created a list of the filenames in Excel. I then created my pipeline in ADF and stored the file names as an array in a pipeline parameter. I first entered a few into the dialogue as shown below:
I then used the code button to enter them directly into the JSON file rather than trusting pasting in all of my names into that little box.
I then created my HTTP linked service that had the URL as my base URL as follows:
Next, I created my HTTP data set.
Note that I have the compression type set to GZip and that the relative URL is using a dataset parameter. I will pass a value down to the copy activity during each iteration of the ‘For-each’ action that will contain the file name that will complete the URL needed to get the file. I store that file name in the data set as shown in the Parameters section of the HTTP data set properties:
I created the target data set to point to Azure Data Lake, nothing out of the ordinary there.
Ok, finally I created the ‘For-each’ copy data activity in my pipeline. I specified that each time it loops, it will iterate through the list of file names included as pipeline parameters by specifying the following in the settings of the ‘For-Each’ activity:
I then created my activity that I wanted my ‘For-each’ activity to run. This would be the ‘Copy Data’ activity that would move the data from the web site to Azure. This is pretty straightforward except when specifying the source. If you remember, I created a parameter on my data set called ‘PipelineFileList’. When you specify that source in the copy activity, because of that parameter, you get a chance to tell it what to store in that parameter as shown:
This is what finally ties it all together. By specifying @item() as the value to pass down to the data set, each time the ‘For-each’ loop iterates, it takes the latest file name and passed it to the source data set, which in turn pulls it from the parameter and uses it as the file name when it makes the request to its HTTP source.
From here, it was just a matter of saving, publishing and letting it rip.
The speed involved was breathtaking. Remember where I specified GZip as the compression method? The pipeline ran for roughly 12 minutes, downloaded, decompressed and copied to Azure Data Lake over 94GB of data.
The next post in the series will be looking at how I start chewing through the data and getting it ready to be pulled into my data model.
I have been working with a retail customer that uses a 445 date table. Matt Allington over at PowerPivotPro wrote a great blog post on what a 445 date table is and how to create one but creating one wasn’t my issue because the customer provided me with the one that they use (Whew).
Instead, I was doing some custom time intelligence with DAX since I couldn’t use the built in functions due to the whacky date table. Specifically, the customer wanted a lot of their metrics expressed in terms of last week, the last four weeks, the last 13 wks and total.
As a side note here, the interesting thing about DAX is that there are usually all sorts of ways to do the same thing. I am not giving a template for how to solve this problem with the following approach, I am just using what I did to show you how I found my first (hah, hah) performance problem and solved it.
I started by defining a simple measure that would give me the latest date for the data that I had in my data set.
CURRENT WEEK ID = MAX ( ‘Date'[DateID] )
Since my Date table only has data for weeks that I have data in my fact table, this gives me the max Date ID for my dataset. With this information in hand, I created some custom helper tables in my model. The example below is the one created for the previous thirteen weeks helper table.
I made a helper table for each of the custom time periods, 13wk, 4wk and 1wk. These tables are created dynamically when the data model is refreshed and they are joined back to the main data table as shown:
So the thinking was this, to make life easy for myself, any time that I had to present a measure over one of these time periods, I would just do the following:
Initial testing went great, it worked as expected and it was super easy to remember this approach.
Of course, I did mention that I when I am doing development, it is against a week of data right? Yeh, probably should have mentioned that. Once I pulled in all of the data, here is a sample of the blazing fast speed that my measure was cranking out:
That is the ‘Server Timings’ readout from Dax Studio. My measure, just one single measure using this approach, took 4.5 seconds to load. For such a simple measure (basically total sales), it actually would have taken almost 15 seconds if it didn’t spread the love out over multiple cores. Even better, its not a lot of data being materialized, the time is all CPU time so I am making the Query Engine thing WAY too hard for such a simple calculation.
When I started digging into the query, I quickly saw my mistake. I don’t quite understand the full details of it because I am not Marco Russo or Alberto Ferrari, but I could see enough from the queries shown in Dax Studio to understand what I did wrong. By specifying an entire table as a filter instead of a column, the Query Engine performed a massive join that pulled in my entire rollup table, the main date table and the other connected rollup tables following the 1:1 relationships. All of the columns in the new mashed up table were then used in such a massive WHERE clause that it actually overran the buffer in Dax Studio before it captured all of the text. Usually DAX studio will cut off the text and tell you how many more lines were remaining but I think I overwhelmed it with my nonsense. The resulting WHERE clause looked like it was filtering on every available value in all of the date columns in all of the joined date tables.
That is only a short piece of the never ending WHERE clause (yes it goes on and on my friend….). Wow, I am not doing something right.
To try and limit the JOIN, I changed my measure to the following, this time specifying a column:
That’s more like it! Total time 48ms. Again, not exactly sure why its so much faster but if you look at the queries, the where clause is much more simpler.
Ok, wrapping up, what did I learn here? Always specify a column in your filter expression when using calculate? Always test your measures against production data? Always keep Dax Studio handy?
I guess so, these are all good things to know. More importantly, I got to immediately put to work the minding bending stuff that I am learning in the SQLBI.com course OptimizingDAX. I am only half way through the course and already am using way more of Dax Studio than I knew even existed before I started. This is not a paid endorsement, btw, I ponied up out of my own pocket to get a bunch of the videos that these two mad scientists cooked up and they have been worth every penny.
Using DAX Studio, I was able to performance test a measure by itself, quantify how bad a job I did with it, see where the slowness was coming from (yes its the WHERE Clause that doesn’t end…), make a change and immediate quantify how fast the measure was once I fixed it.
I will try to blog more perf scenarios as I come across them but for now, if you don’t know DAX Studio, and you are responsible for DAX/Power BI/SSAS Tabular performance, you need to learn it now.
And in case I didn’t firmly stick Lambchop into your ear with this post, enjoy: