Content in Tableau Community Forums

Anyone seeing Tableau 8 crash on Select Shape Palette?

tableaucommunity@tableau.com — Thu, 24 Oct 2013 14:53:24 GMT

We've had two separate installs of Tableau 8 start crashing when, in a worksheet with "Shape" type marks, selecting the "Edit Shape..." dialog from the shape pane, we try to change the selection in the "Select Shape Palette".

I just wanted to see if anyone else was having similar issues. We can assign shapes from the default palette with no issue. Our installs were upgrades from Tableau 8 Beta, but the problems didn't happen immediately -- we've published packages with various shapes prior to the crashing situation.

We are not using custom shapes. The crashes are 100% recreate-able, every time, on any worksheet, once they start.

The SOLUTION we've found is to re-install Tableau 8 (or just select the "Repair" option from the installation). This solves the problem.

Still, I wanted to post to see if it was endemic or something we may be doing wrong, or something Tableau may want to open a bug report on.

Thanks!

---Chip

Filtering Data by Qualifying Events

tableaucommunity@tableau.com — Thu, 17 Oct 2013 19:29:31 GMT

Welcome to today's post where I ask a question that I think is difficult and someone gives me a fantastic reply in, say, 30 minutes.

I'm struggling with a problem in healthcare data. I have a large dataset of members with claims with specific dates and diagnoses. My users want to explore how members with particular conditions fare over time.

For example, we'd like to split the population into people who ever had Asthma or Schizophrenia or Toe Fungus vs. people who did not, and compare the number of claims, severity, cost, diagnoses, and procedures across those sets. There are thousands of diagnoses, so building a separate flag for each possibility is arduous. I'm looking for a better way.

I've attached an example using Tableau's sample "Orders" data which almost-but-not-quite works. Here I use the "Product Sub-Category" as the qualifying event. That is, the user can select any Product Sub-Category, and the data set is divided into Customers who have ever purchased something in that Sub-Category (Qualified), and those who have not (Non-Qualified). Except the "Qualified" group generally appears as an asterisk (undefined). This makes sens, I just can't quite figure out how to get around it.

In the attached example I have "Paper" selected as the user-selectable sub-category. This is an important example because there exist customers who have never bought Paper, customers who have bought paper and something else (ever), and customers who have ONLY bought Paper. All three groups must be accommodated. I don't mind if the latter two are combined (as "Qualified"), but in my example they are not, since the attribute is undefined ("*") for Customers who had both.

In a perfect solution, I could base the Qualified flag on a selection of multiple things, for example Product Sub-Category, Product Name, or Total amount spent (i.e. separate my big spenders from little spenders), but I'll take it one step at a time.

Some oft-requested tableau features would help, such as passing the parameter to custom filter or SQL, so I could calculate only one value (Qualified/Non-Qualified) for each Customer. But can anything be done with the Tableau 8.0?

Many thanks!

---Chip

Can I synchronize "Measure Names" Filters?

tableaucommunity@tableau.com — Thu, 20 Jun 2013 20:40:18 GMT

I've attached a very simple workbook. It's a dashboard with two sheets (this is using Tableau's sample data, mind you). The sheets each display simple crosstabs. My users want to select which metric will display. I can get the "Measure Names" filter working fine, but I can't synchronize it across worksheets on the dashboard.

Any ideas?

Key to the issue is keeping the formatting correct... some measures are dollars, and some are counts (integers), while others are fractions (real/decimals). If I create a calculated field to pull the value from a parameter (rather than use the Measure Names shelf), I can't consistently format it properly.

Either approach, or another is welcome. I'm open to suggestions!

Many thanks,

---Chip

Slow performance from ISNULL in WHERE Clause

tableaucommunity@tableau.com — Fri, 12 Apr 2013 20:44:11 GMT

NO question here, I have a workaround... I don't know why Tableau does this, but I wanted to share since I couldn't find another reference to the problem...

First, I'm using Tableau 8.0 on SQL Server 2012. No idea if this happens anywhere else...

I have a calculation called "Group or Provider Name": IFNULL([GROUP_NAME], [PROVIDER_NAME])

I use it fairly frequently in charts... basically if someone is part of a group, we want to show the group information, otherwise we just show their individual information. No worries and this works great most of the time.

Sometimes, though, and this happens more frequently when running on the web server rather than on the desktop version, Tableau will basically hang on a report that is normally a 15-second report. I've waited 15+ minutes with nothing.

The REASON is this: Tableau submitted a query to SQL Server doing this:

SELECT TOP 1 ISNULL([SPGA].[GROUP_NAME], [SPD].[PROVIDER_NAME]) AS [Calculation_9860412105917202]

FROM [Facts].[CLIF] [CLIF]

INNER JOIN [Dimensions].[SPD] [SPD] ON ([CLIF].[SERVICE_PROVIDER_T1_SKEY] = [SPD].[PROVIDER_T1_SKEY])

LEFT JOIN [Dimensions].[SPGA] [SPGA] ON ([CLIF].[SERVICE_PROVIDER_T2_SKEY] = [SPGA].[PROVIDER_T2_SKEY])

WHERE (ISNULL([SPGA].[GROUP_NAME], [SPD].[PROVIDER_NAME]) IS NULL)

I added some bold and underline for emphasis... Tableau is requesting a single record back using the ISNULL function (not IFNULL, which I used in my calculation -- subtle important difference). Apparently Tableau wants to check if the result set will ever be NULL (note the IS NULL [with space] in the where clause after the ISNULL [no space] function). This single query destroys my performance.

The SOLUTION, is to ensure that there is a default value and that the final result of the IFNULL can never be NULL... that is, I changed the calculation to:

IFNULL(IFNULL([GROUP_NAME], [PROVIDER_NAME]), 'Unknown')

So it will return 'Unknown' instead of NULL if both names are missing (a circumstance which never happens in the dataset, by the way). Now things are fast again. I don't know if Tableau stopped running the silly query entirely or just rewrote it so SQL Server can just optimize it better by shortcutting the WHERE clause, but in either case performance is good again.

Thought you all should know.

Dimension calculation based on MAX() subquery?

tableaucommunity@tableau.com — Wed, 02 Jan 2013 14:23:53 GMT

In a health insurance database, we have Members who may or may not have Claims. I have a parameter on a slider that is "Limit Number of Months", an integer, between 1 and 24. It limits my dashboard to Claims with a Service Date in the last however-many months.

So far so good. I have a handful of charts with various dimensions that all filter by this successfully.

Now, I want to add a chart with the number of Members -- COUNTD(MEMBER_ID) -- in a stacked bar arrangement where members WITH a claim in the parameterized time period get one color (we call this "Utilization") and members WITHOUT a claim in the time period get another ("No Utilization").

Here's the calculation for "Utilization Flag" (it was a boolean, but ignore that for now):

IF ISNULL(MAX([Service Date])) THEN 'No Utilization'

ELSEIF MAX([Service Date]) >= DATEADD('month', [Limit Number of Months]*-1,TODAY()) THEN 'Utilization'

ELSE 'No Utilization'

END

MAX() forces this to be a Measure, not a dimension, which I understand, but if I take the MAX() out, then I have a situation where some Members are double counted, since the [Service Date] is tied to the Claim. That is, a Member can have a Claim in the service period and outside of it, so they would be double counted if I counted distinct MEMBER_IDs grouping by this Utilization Flag.

So I need the MAX(), since if the most recent claim wasn't in the period, then the older ones don't matter. Now, though, I can't stack the bars because my boolean result is now a Measure.

The best I can do is to set the Level of Detail to MEMBER_ID, and make Utilization a column header. This works visually (except I'd prefer a true stacked bar), and the numbers are correct, but now if I mouse over it selects every individual member... I have a uni-color stacked bar with thousands of unit-width stacks (note the high number of marks in the screenshot)! Also, this makes it impossible to calculate and display the "total" number of members.

Lastly, I'm reticent to add another subquery to the process, although my one possibility is to add "latest activity" as a field in the database, but I feel like I shouldn't have to. The dashboard is filter and drill heavy, so having a single data source is a huge boon.

I can't publish data due to privacy concerns, but I've included a highly aggregated screenshot.

Any ideas are welcome!