Currently cross tables are presented with a default sorting, as alphanumeric ascending in precedence of column/row positions. This default can now be adjusted to sort based on the results, but not on the values in the headings themselves.
Cross tables would be much more powerful with the following improvements to sorting:
1) Options for sorting by row and column headings, defined individually for each row/column header, as:
a) Ascending
b) Descending
c) Numeric/Date Ascending (where number/date field type used as a category)
d) Numeric/Date Descending (where number/date field type used as a category)
e) by non-displayed "sort key" column
f) manual re-arrangement to required order (e.g. by drag and drop)
2) Ability to define hierarchy of sorting precedence - not just based on column position
This would allow the cross table to present data in a logical order with similar / sequential categories shown in an order which makes sense and allows easer interpretation of the data.
This is most important for Analyst and Business Author licences, but adjustment would also be useful for Consumer.
Example to show how the same data can be sorted in different ways
Raw Data
reference |
category |
colour |
colour sort |
result |
A01 |
30 |
red |
1 |
5 |
C02 |
100 |
orange |
2 |
2 |
D01 |
10 |
yellow |
3 |
5.5 |
D01 |
5 |
green |
4 |
9 |
D01 |
75 |
orange |
2 |
3 |
B01 |
100 |
indigo |
6 |
1.5 |
B03 |
30 |
orange |
2 |
2 |
B02 |
75 |
blue |
5 |
12 |
A01 |
100 |
blue |
5 |
11 |
cross table data
Example 1
column headers: sort by reference (ascending), category (numeric ascending). Row Headers: sort by colour sort.
reference |
category |
colour |
|
|
|
|
|
|
|
red |
orange |
yellow |
green |
blue |
indigo |
A01 |
30 |
5 |
|
|
|
|
|
|
100 |
|
|
|
|
11 |
|
B01 |
100 |
|
|
|
|
|
1.5 |
B02 |
75 |
|
|
|
|
12 |
|
B03 |
30 |
|
2 |
|
|
|
|
C02 |
100 |
|
2 |
|
|
|
|
D01 |
5 |
|
|
|
9 |
|
|
|
10 |
|
|
5.5 |
|
|
|
|
75 |
|
3 |
|
|
|
|
Example 2
column headers: sort by category (numeric ascending), reference (ascending). Row Headers: sort by colour sort.
reference |
category |
colour |
|
|
|
|
|
|
|
red |
orange |
yellow |
green |
blue |
indigo |
D01 |
5 |
|
|
|
9 |
|
|
D01 |
10 |
|
|
5.5 |
|
|
|
A01 |
30 |
5 |
|
|
|
|
|
B03 |
30 |
|
2 |
|
|
|
|
B02 |
75 |
|
|
|
|
12 |
|
D01 |
75 |
|
3 |
|
|
|
|
A01 |
100 |
|
|
|
|
11 |
|
B01 |
100 |
|
|
|
|
|
1.5 |
C02 |
100 |
|
2 |
|
|
|
|
Example 2 would be useful because I often use cross tables without hiearchies on the vertical axis. I use the labels to display some columns and in the cell values I display data which has to be further splitted horizontally. So basically, like a table with an attached cross table and unique values in the first few columns. In that case, being able to sort would be very useful, because rearranging the labels would be rather tedious and destroy the layout of the table.
+1 on this. Looks like a year has past on this one with no closure. I think being able to sort the cross table vertical headers would go a very long way. If the vertical header is a date, it would be nice to have the most recent date at the top, not the bottom. Also sorting the horizontal headers in the same fashion would be great as well. Thanks in advance, Ben
Louise - sorry for the delay but please email me at nambernt@tibco.com to setup a call to look at some of your usecases.
Hi Niklas,
A) having independent Ascending/Descending sort definitions for row and column headers would help, but wouldn't allow us to support all our users' requirements.
The other key point is to be able to define the order, particularly for the horizontal row headers, using a sort key. My example here was colours of the rainbow, since it's easy to understand why you may want the colours to be presented in an order which is different to the alphabetical display – however, our real life examples relate to presenting different methods and result types in a particular logical order. My understanding is that this could be achieved by adding the sort key to the defining headers, but there is already limited space in the way cross tables are presented – perhaps an alternative solution for this would be to allow row and column headers to be included in the cross table definition, but hidden in the output. Without adding sort key rows & columns, one of our typical cross tables might be defined by 8 column headers and 5 row headers.
B) We do have examples where our Users want to present the row fields in a particular order, but sort by a different precedence. One reason for this might be that they want certain information to be closely associated with one another within the headings to make it easier to read, but also, when they are viewing the data with different sort precedents, they want the column fields to be in the same order, making it easier to compare the data.
I hope that helps - if you PM me perhaps we could set up a short meeting where I can show you some real life examples of the type of report/visualisation we are trying to make available via Spotfire.
Many Thanks
Louise
Hi Louise! I am reviewing this idea and would like to ask a few questions.
A.) if we had the capability to sort cross table vertical headers either Ascending or Descending, this would go a long way right? (today the vertical headers are always sorted based on the columns sort order). And I understand you also want to have the capability to specify sorting of horizontal headers (locally in the cross table, as opposed to the column sort order).
B.) In your example 2 above, you have kind of inverted the hierarchy by sorting first on Category. You could acheive the same result (analytically, but not visually) by just moving category to be the highest level of the hierarchy. Is there any specific reason that you would want exactly the visual representation you suggested in Example 2? For me it seems more natural for a crosstable to have its highest level of teh hirerarchy on the left-most position. Looking forward to hear your comments! Kind regards - Niklas