When the user clicks on the cross icon, the corresponding block should be removed from main view and filter view.

The user should only be able to switch to meaningful other visualizations that make sense for the data type.

For matrix, numerical and categorical blocks, we can set the default vis to a heat map.

Transition to item level: When the label column is added and the items on individual level fit the screen, the table will be expanded to item level. If it doesn’t fit the screen, the top ten items will be expanded and the rest of the table will stay aggregated. The number of expanded items can be adjusted by brushing on label column. To aid brushing we will support snapping to multiples of 5.

The filter should perform intersection when we select different types of filter.
For example: If we select the male from one categorical data and US from another categorical data. The result should only display excluding male and US.
[1,2,3,4]- Indices of Male
[3,4]- Indices of US
The result of intersection should be [3,4].

Assume there is already some kind of filters active. Now if the user adds the new attribute, the new attribute data will be filtered based on the current filtering criteria.

Synchronize height of the rows in separate blocks.

Should look similar to sketch:

The categorical data will be represented by different color encoding. When we select the category the opacity will be reduced. The rest of the remaining category are selected for filtering the data.

What should happen, if we delete a column where we have filtered something out in the corresponding block in filter view?
e.g. columns of gender, career status ...we filter out inactive artists, then delete the career status column.

• A) the vis should come to the state before the career status filter was applied
• B) keep the filter - but currently we lose track of it, because the column with the header is deleted and the corresponding block thus as well...(and currently - if we add the career status again, the filter view has both categories as selected, not inactive grayed out)

Even though after some playing around I have reset the numerical filters to cover the whole range, some values god missed out

• it is currently hard (sometimes impossible) to include boundary values again after you previously filtered them out
• it is impossible to select only small range in the middle of the range of values - the other handle of the slider starts moving because of the size of the handle
• reimplement this with d3 brushing, it's nice and easy to interact with

needs to be implement for each data type manually. Since avaialable visualizations aren't designed for user interface.

It consists block which will be useful for changing the main view on changing the selection.

The column manager should assign new columns for each dataset a user selects. The manager should also take care of assigning space to the columns. The columns itself (own class) should then decide on its own how to fill this space and how the data needs to be aggregated given the assigned space.

@sgratzl will create an initial version where @katush and @mijar can then continue with working on the different features.

Draw the lines connection between different data attributes as shown in figure.

Should look like in sketch if unaggregated.

Filter Data excluding Barbados

Color coding is missing in TCGA data for categorical data

As a start, I suggest to let the blocks within the columns always fill the vertical space they have. In the current implementation, they are sometimes shown in full height, and sometimes they are only 100px height.

will manage one dataset and is responsible for

• toolbar actions: search, sort, delete
• visualizing the component using MultiFormGrid
• apply given sorting

Columns should be sorted and stratified hierarchically. The sorting order is driven by the order in which the filter widgets appear on the right side (see #39).

In the current implementation all columns have the same width. As a starting point, the width of the column should be set according to the block width. The goal is to get rid of the white space between the blocks in order to see an actual table.

The filter for matrix type of data.

Register filter dialog only once for each attribute.

Introduce and synchronize sorting of items in blocks.

For now, only the colors for gender are set. All categorical attributes should get some color. Makes it also easier to see if filters work.

The header of the column should be invisible by default and should appear on demand.

If matrix is added for the first time, the section is created...if we want show the same matrix for the second time (e.g. with different stratification) the corresponding section in Filter View is not yet created

Current State:

Remove multiple line with scaled thickness.

The first user will choose the different IDTypes (like Artist or Countries)
Then if there are two idtypes there will be two filter manager for each idtypes.
There will be dropdown selection for each idtypes which will contain all the data it holds
For example in Artist: the data could be artist name, gender, origin
In idtype of countries, the data could be continent, sold albums

When you add an attribute twice, it appears only once in the fiter view, which is correct - the problem is when you want to delete one instance of this attribute in table view and keep the other - the corresponding block in the filter view disappears.

1. Add origin
2. Add Career status
3. Add origin again
4. Delete one of Origin columns from Table View

acts as a container and manager for the invidiual filters

I would suggest the following behaviour:

• the size of the handle does not limit the user in the selection of desired interval
• once moved somewhere in the block, the handles do not change their position by themselves, only by interaction
• moving some handle does not affect the position of the other
• maybe use gray handles and change the color to black and status to draggable on hover to allow selection of even tiny intervals?

I guess we should also allow to create discontinuous intervals, something like this:

Loading TCGA data for one of the tumor types is important because it gives a bigger dataset to play with during development. Furthermore, this will also be the dataset that our collaborators will use to provide feedback.

The numerical value should be represented in linear gradient and with brushing for the selection of the range.

• Scale the column to keep row correspondence, not strings
• Keep changes when vis updates

• allow dragging
• automatic change of order when the user clicks on the sorting icon in toolbar of a column in the table view (this attribute becomes 1st sorting criterion - moves to the top of filter view)
• Trigger the sort event after dragging finishes.

• circle visualization similar to bar plot
• circle visualization showing an aggregated version (e.g. mean)
• single heatmap cell visualization (e.g. mean)

The user should be able to resize the width of the column by some icon on the header or by dragging left or right as in lineup.

Should work with numerical and categorical attributes.

The user will be able to input string and update the main view accordingly.

Empty selectiion

If we select as in filter dialog exclude female , inactive and artists starting with G then we have empty selection as below.

Remove all blocks from filter and main view. Or what should we do ?

If the attribute is String type. The user should be able to filter the data with string search pattern