Plots are sometimes too small, particularly when inspecting scatter plots of large datasets. It would be convenient to have the option of resizing the plot grid. This resize could work in a simple manner: when hovering the mouse on the corner of a plot for more than a set time, then to guide lines (horizontal and vertical) appear. The user can then adjust the (rectangular) size of the plot by dragging these lines.

Right now the login form/error dialogs are very basic

To select single values or invert selection in the categorical selector, I find SHIFT + mouse click easier to use than long presses.

Firstly, I'd like to contribute to this project as I have some ideas in mind however trying to compile and run mirador v1.3 and v1.4 in both Eclipse and Intellij has both lead to disappointing results. If you are going to ask if I have included the necessary .jar files, yes I have but it keeps throwing up this error and would like your help on it.

When a sorted variable is automatically resorted (for example when changing a range or the P-value), the profile plot is not updated.

Make sure that points in scatter are properly sized, etc.

The clipping is not properly handled by the pdf recorder, this might need to be reported in the processing repo though.

It would be very useful to being able to switch the histograms (in case they are made conditional) and eikosograms between the modes Y|X (current) and X|Y, otherwise the user needs to witch row/columns around, which takes additional time

When a range was set in the previously dataset (bc the ranges were not reset when loading a project)

Is it possible to add a new data source (such as a database)? Where should I begin to look?

After selecting an included example data file nothing happens and no windows are shown, CPU idle. Java throws:

Exception in thread "Animation Thread" java.lang.NoSuchMethodError: javax.media.opengl.GLDrawableFactory.initSingleton()V
at javax.media.opengl.GLProfile.initProfilesForDefaultDevices(GLProfile.java:1696)
at javax.media.opengl.GLProfile.access$000(GLProfile.java:77)
at javax.media.opengl.GLProfile$1.run(GLProfile.java:186)
at java.security.AccessController.doPrivileged(Native Method)
at javax.media.opengl.GLProfile.initSingleton(GLProfile.java:172)
at javax.media.opengl.GLProfile.getProfileMap(GLProfile.java:2193)
at javax.media.opengl.GLProfile.get(GLProfile.java:942)
at javax.media.opengl.GLProfile.getGL2ES1(GLProfile.java:833)
at processing.opengl.PJOGL.initSurface(PJOGL.java:254)
at processing.opengl.PGraphicsOpenGL.initPrimary(PGraphicsOpenGL.java:6324)
at processing.opengl.PGraphicsOpenGL.requestDraw(PGraphicsOpenGL.java:1658)
at processing.core.PApplet.run(PApplet.java:2259)
at java.lang.Thread.run(Thread.java:745)

Using OSX 10.9.4 on MacBookPro9,1, NVIDIA GeForce GT 650M/Intel HD Graphics 4000
Java 7 Update 60 installed

The plots of any variable against itself is a regular histogram, but they are easier to confuse with eikosograms as there is no explicit indication of the difference, specially for categorical variables with few values.

When selecting a subgroup of categories, the labels in the X axis for the variable being used as the range appear to be wrong.

the default missing value is currently the question mark "?". Perhaps it is better just to use the empty string, or always consider missing value those which are empty, regardless of the missing string specified by default or in a project file. Right now it seems that an empty string is regarded as a valid category.

A few times this error happened when searching a variable with the search bar"

Exception in thread "Thread-64" java.util.ConcurrentModificationException
at java.util.HashMap$HashIterator.nextEntry(HashMap.java:926)
at java.util.HashMap$KeyIterator.next(HashMap.java:960)
at java.util.AbstractCollection.addAll(AbstractCollection.java:341)
at java.util.HashSet.<init>(HashSet.java:117)
at mirador.app.SearchBar$SearchResults$SearchTask.run(SearchBar.java:421)

Does not result in all the other variable rows closing.

Right now dragging (and arrow keys) is the only way to move through rows and columns. Scrollbars will make the interface more intuitive, as it is not clear to everyone that dragging or keys control movement.

The green triangles that indicate that a variable pair is correlated at the current significance level should appear gradually, with a soft animation indicating that they dependency test is being conducted.

And see selection in other plots (akin to what glueviz does).

When the group has thousand of variables, clicking on the columns checkbox results in the UI hanging for a while, while Mirador traverses the tree to select/deselect all variables.

On Windows 7 a "Couldn't create a writer" error is recieved (see https://twitter.com/Johan/status/663759898262548481) (Mirador beta 1.4)

Initially the package was deployed under \programs, later I tried to move the package to another folder.

Might be that the \programs folder is "not open enough" to write for the package, but even after the move it keeps giving the same path (see screenshot https://twitter.com/Johan/status/663759898262548481)

Is the program path set in the registry at 1st start? If som how can this be reset?

Right now it's using one std around mean (for numerical dependent variables), and two std around mean, should use the 1st and third quartiles for the central box (and indicate the median instead of the mean) and probably Q1 - 1.5 x IQR and Q3 + 1.5 IQR for the whiskers:

http://en.wikipedia.org/wiki/Box_plot
http://en.wikipedia.org/wiki/Interquartile_range

Empty plots (for example when one variable only takes one value) should show a message explaining this situation, otherwise the user won't know why they are empty.

alt/option is the standard key for alternate options in Windows/Mac UIs

This might need some extra work to write the appropriate startup scripts and test on a couple of distros.

This allows to do comparisons with a variable with lots of missing data... however a way to accomplish this without any addition is by setting the "Available data" option to don't use.

So the user can see what went wrong when loading a file

When the profile view is open, and the P-value is changed, the plot is not updated, although the variables are resorted.

This is useful to focus on a specific plot/group of plots.

Searching covariate, row, or column in the search box slows down the interface significantly for large datasets like NHANES with more than 2,000 variables

Unstable and has image display issues on Windows 7 Home Premium nor Professional, with newly installed OSes and different hardware configurations.

With Win 7 Home Premium on Toshiba Satellite series, and with Win 7 Professional on both hardware (ThinkPad W series) and VM either crash upon loading example data, or displays patches and scratches of black and white blocks and stripes.

Was wondering if Mirador be available for Linux distributions(esp. Ubuntu/Debian) soon? Would definitely appreciate further info on the progress! Alternatively, any fixes and enhancements that could resolve the stability issue on Windows will be immensely valuable. Thanks much!

When scrolling horizontally very quickly to the left, so the plots are closed, the range selectors remain visible for a few seconds longer:

As the user examines the data, the significance test should be adjusted to reflect the increasing number of dependency hypothesis being tested.

Merge MiraApp and Mirador classes, the initial open file dialog could be shown with the options panel already visible.

So far, I think the marginal histogram is not particularly useful. A conditional Y|X, X|Y histogram would be much informative, to complement the eikosogram specially in cases which the bins have very different populations.

The project's .mira file is empty after generating exported dataset using the export selection option in the profile view

Exporting the selected profile variables can hang the UI, specially for large datasets. This probably needs a separate thread...

Right now there is no notification when the export of a data set (either from the profile or the main view) concludes.

So that Mirador opens in the same state it was left at.

Hello. I downloaded your source code and attempted to run the application but it looks like I am missing both "processing-core" and "processing-pdf" projects to complete building the project. I found "miralib" which you uploaded. Would you be able to tell me where to find the others?

Using the hepatitis example dataset

In addition to range reset, it would be useful to reset sort configuration and column selections for an easily accessible button.

I have downloaded the windows version and unzip, double click to run. I am running on a Windows7 laptop.

When hovering over a point in the scatter view, it should be highlighted across all visible plots

Instead of png, better for publication, editing.

Sometimes it is possible to move the plot type selector to an inexistent position after the eikosogram option. This results in the following error:

Unsupported view type: 3
java.lang.RuntimeException: Unsupported view type: 3
    at mirador.views.View.create(View.java:52)
    at mirador.app.RowPlots$Plot$1.run(RowPlots.java:227)
    at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:471)
    at java.util.concurrent.FutureTask.run(FutureTask.java:262)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1145)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:615)
    at java.lang.Thread.run(Thread.java:744)

It is useful to display count/total for each highlighted area in either the scatter or the eikosogram, next to the percentage.

The P-value from the previous session remains after loading a new project (w/out restarting the app)

When loading large datasets (+10k rows), Mirador could create a random mask so only a smaller subset of the rows are consistently used throughout calculations and visualizations.

When changing the name of the selection, the prefix for the dictionary file is not updated, also the last folder used to export the files is not remembered.

The order of columns is X,Y in the exported table, maybe Y,X would be better since that's the order in the export profile?