Concepts
********
Filtering and display
---------------------
XPLOR allows users to easily visualize their multi-dimensional data. Its
base principle is to apply filters on the data to select a sub-portion of it
that will be displayed. In XPLOR terminology, *data* refers to the original
data, while the displayed sub-portion is called a *slice*.
**Steps in using XPLOR**
.. image:: image/globalinterest.png
:align: center
:alt: XPLOR: filtering and display
XPLOR window
-------------
The organization of the XPLOR window reflects this principle as it is composed
of a control zone on the left, where the user selects the filters to
apply; and of a display zone on the right, where a number of interactions are
possible such as reorganizing the layout by moving the labels' accross axes,
zooming, etc.
**Screenshot of the Window with the control zone and the display zone**
.. image:: image/windowExp.png
:align: center
:alt: Screenshot of Window
Labelled data
-------------
To appropriately display labels and scales, XPLOR needs to know what the
data values, and what each data dimension, represents (time, space, unit,
etc.). Any input data is therefore converted first to an xplr.XData object,
which is a container for the N dimensional data and all the relative metadata.
**Example of a Xdata instance**
.. image:: image/xdataexample.png
:align: center
:alt: Illustration for an xplr.XData object
Filter object
-------------
Filters apply to one or several specific dimension(s) of a labelled data (i
.e. of an xplr.XData object) and return a labelled data.
In the example below:
- the initial *data* has 5 dimensions: 'day', 'country', 'type', 'data'
and 'norm.', its size is (168 x 200 x 2 x 2 x 2)
- 4 filters are applied to dimensions 'country' (4 countries selected),
'type' (1 value selected), 'data' (2 values) and 'norm.' (1 value)
- the resulting *slice* still has 5 dimensions: 'day', 'country', 'type', 'data'
and 'norm.', but its size is reduced to (168 x 4 x 1 x 2 x 1); only 3
dimensions have more than one value: 'day', 'country' and 'data'
- this *slice* is displayed in the display part: dimension 'day' is
located in x and is used to form signals (it is called an *internal
dimension*), dimension 'data' is dispatched vertically and dimension
'country' is dispatched as a 2x2 grid (they are called *external
dimensions*).
**Screenshot of the Window showing the filter applied**
.. image:: image/filter_example_coronavirus.png
:align: center
:alt: Example XPLOR window with active filters
Zoom filters
--------------
The *slice* can be furthered reduced by zooming into it. It is possible to
zoom in either by dragging a rectangle with the mouse left button, using the
scroll wheel, or manipulating the sliders on the display side (see
screenshot below). The result of zooming is called the *zslice*, it is
again an xplr.XData object.
.. image:: image/filter_example_intrinsic.png
:align: center
:alt: Example XPLOR window with active filters and zoom filters
Slicing chain
--------------
The diagram below details the steps of applying filters and zoom filters to
an original *data*, and shows some of XPLOR's internal objects being involved.
It follows the example of the previous screenshot:
- The original *data* has 5 dimensions: 'x', 'y', 'time', 'condition' and
'repetition'
- 3 successive *filters* are applied respectively on dimensions 'condition',
'repetition' and 'condition'. Each intermediary step is stored in a distinct
xplr.XData object, together they form the so-called *slicing chain*,
which is stored inside a new XPLOR object: the *slicer* (this object is
responsible for applying filter operations each time the original data or a
filter has been modified). The result of this slicing chain in the *slice*.
- This *slice* itself is the initial element of a second slicing chain
handled by XPLOR's *zoom slicer* object: 2 *zoom filters* are applied,
respectively in the 'x' and 'y' dimensions. The final result is the *zslice*
that is displayed (as in the figure above).
.. raw:: html
Synchronizing windows
--------------------------------------------------
The diagram below now shows the implementation of a more complex situation
with two lists, a spatial display and a temporal display of the same original
data.
- The lists filter the 'condition' and 'repetition' conditions for both
displays.
- In the spatial display, one can either move the cross or draw
spatial regions of interest to control the 2D filter applied in dimension
'x-y' to the data appearing in the temporal display.
- And vice-versa, in the temporal display one can either move the cross or
draw temporal regions of interest that control the filter in dimension
'time' applied to the data shown in the spatial display.
Beneath each display are shown the filters *controlling* what is shown in
the display (they appear in the slicing chain and zoom slicing chain already
commented above), but also the filters *being controlled* by the lists and by
the displays.
All colored elements in the diagram are filters and zoom filters. A single
filter (or zoom filter) object appears at several places in the diagram.
Indeed it appears both where it is *controlling* and where it is *being
controlled* (and it can be controlling in several displays, as is the case
for the 'condition' and 'repetition' filters; it could also be controlled in
several displays though there is no such case here). In addition it also
appears in a new XPLOR object that we introduce here: the *bank*, which
keeps pointers on every existing filter and zoom filter.
Finally, red texts and arrows show an example sequence of events occuring when
the user performs an action in a display: here select a new region of interest
in the spatial display.
.. raw:: html