iFit: iData object description

Commands we use in this page: iData, get, set, setalias, getalias, setaxis, getaxis, labels, methods
See also: iFunc, Fit, Math, Plot, Load, Save

The iData class is a structure which holds some multidimensional data set (numerical blocks), as well as additional metadata and axes definitions, Signal, Error bars, Monitors, and user defined fields (aliases). The instantiated objects, which are used as normal Matlab variables, often contain 1D, 2D, 3D, nD histograms or nD event lists. You will learn below how to customize these objects (set, get, setalias, getalias, setaxis, getaxis, indexing, ...).

A large number of methods can be used with iData, including data import, plotting, saving, fitting and mathematical operators.

For impatient readers, we advise you to refer to the quick Setting Signal and Axes documentation in the Load page, in order to customize the object. For the others, just go on reading how objects are built, and how to handle them.

Derived classes

The iData data sets also exist in a set of specialized flavours (sub-classes) which inherit full iData capabilities, and add a few more specialized functionalities:

Introduction to the iData objects

An iData object is obtained for instance by transforming a file name into an iData object a with
>> a = iData([ ifitpath 'Data/ILL_IN6.dat'])	% imports a file into an object
a = iData object:

[Tag] [Dimension] [Title] [Last command] [Label]
tp460929 [257 512] 'File ILL_IN6.dat ILL Data (n...' tp460929=load(iData,...
which results in an iData object. An other way to create such an object is to import directly Matlab variables:
>> a = iData(rand(10))
>> a = iData(x,y,signal) % give the axes first, then the Signal
>> a = iData(gauss, 5, -5:5) % a gaussian with width=5, evaluated along axis -5:5
These objects, which internal structure can be obtained with the struct(a) method, hold a number of important items:
the iData object
          internal structure
The iData object internal internal structure.

In principle, the objects should not be accessed or modified as structures. Indeed, some dedicated methods, which match the usual Matlab commands, have been designed to maintain the object integrity along operations. This includes methods for
In the following, we shall present how to manipulate iData objects, enrich them with additional information, define relations between the object parts as Aliases, define the object Signal and its associated Axes.

Any object dimensionality can be handled. A more limited support exists for event based data sets.

To get more help: list of most iData methods with comments

Inquiring the iData object (get, disp, display)

The main idea here is that iData objects can be handled the same way as structures and Matlab Graphics Objects using get and set methods.

A quick overview of an iData object a is displayed when issuing the command
>> a = iData([ ifitpath 'Data/ILL_IN6.dat' ]);	% converts a file name into an iData object
>> a
ans = iData object:

[Tag] [Dimension] [Title] [Last command] [Label]
tp481575 [257 512] 'File ILL_IN6.dat ILL Data (n...' tp481575=load(iData,...
showing the object unique ID, its dimensions, title, the last command which produced the object, and an optional label (here empty). The same output is given by the display(a) command.

But a more useful formatting of the same information is obtained using get or disp
>> get(a)
>> disp(a)
which both result in the following output
a = iData 2D object of size [257  512]:
               Title: 'File ILL_IN6.dat ILL Data (normal integers)'
              Source: 'iFit/Data/ILL_IN6.dat'
             Command: {'tp481575=load(iData,'iFit/Data/ILL_IN6.dat');'}
            UserData: ''
               Label: ''
         DisplayName: ''
             Creator: [1x119 char]
                User: 'farhi on localhost'
                Data: [1x1 struct]
                Date: '20-nov.-2007 23:11:26'
                 Tag: 'tp481575'
    ModificationDate: '12-Oct-2010 17:48:16'

Object aliases:
          [Name]                           [Value] [Description]
         Signal  Data.IIIIIIIIIIIIIIIIIIIIIIIIIIIII_13   Data Signal [0:27172]
           Error                      sqrt(Signal)   Error on Signal
        Monitor                                 1   Monitor (weight) '1'
     Attributes                   Data.Attributes   Headers from ILL_IN6.dat

This latter display indicates the definition of the Signal, Aliases and Axes (if any - see below).

The object 'data' is stored in the 'Data' field of the object structure. This is usually were any file is imported.
As we shall see below, it is simple to extract the object value (Signal) and Axes:
>> a.Signal					% returns the Signal value
>> a{0} % returns the Signal/Monitor value
>> a{1} % request the 1st axis values
A single field can be accessed in an object or an array of objects, by means of the get function (just like for the Matlab graphics objects):
>> a=load(iData, [ ifitpath 'Data/*.scn'])	% returns an array of iData objects
>> get(a,'Title') % returns the Title of all objects in the array
>> get(a(1),'Title') % returns the Title from the first object
where this latter is equivalent to accessing
>> a(1).Title
>> b=a(1)
>> b.Title
Field names are case insensitive, that is b.title will also work.
Any field may be accessed, including subfields, just as a normal structure
>> get(a,'Data.VARIA.A1')	% returns all A1 values from the array of objects
>> a(2).Data.VARIA.A1 % same but only for the second array element
>> a(3).Data % returns the complete Data structure from the 3rd object in the array
Accessing subfields requires a case sensitive name match, that is a a(3).data works (single field accessed), but a(3).data.varia.a1 does not (requires fully qualified path to the data).

The get function can also apply with iData arrays, as shown above, and then returns cell arrays.

The content of objects can be searched with the findstr and findfield methods which search for a character string as a content or as a field name:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> [match, field]=findstr(a,'TAS');% searches matches within object contents
match =

'UPt3;File sv1850.scn ILL TAS Data'
[1x105 char]
[1x169 char]
'ETAS= '
field =

>> findfield(a,'TAS') % searches matches within object members/aliases/fields
ans =

Some data file formats provide additional 'attributes' to each numeric content (e.g. NeXus/HDF/NetCDF/CDF. See Loaders). You can access these associated attributes with the fileattrib method:

>> fileattrib(a, 'Data.Attributes.PARAM.ETAS')
ans =

>> a=iData([ ifitpath 'Data/IRS21360_graphite002_ipg.nxs' ]);
>> getalias(a,'Signal')
ans =

>> fileattrib(a,'Signal')
ans =
signal: 1
axes: 'axis1,axis2'
units: ''
unit_label: ''
The findobj method can be used to inquire all iData objects available in the workspace:
>> a = findobj(iData);
>> a = findobj(iData, 'Property', 'Value'); % searches for objects that match a Property/Value pairiData plot 2d: plot(a)
It is also possible to plot the object, and we direct the reader to the Plotting page.
>> a = plot(a);
Last, the object  Signal normalized to the Monitor can be displayed as a Table/spreadsheet (beware the size of the object !)
>> a = edit(iData);Edit(iData)

Modifying the iData object (set)

The content of the objects can be modified using the set function:
>> a=load(iData, [ ifitpath 'Data/*.scn'])	% returns an array of iData objects
>> set(a,'Data.VARIA.A1',40); % modify the whole array Data content
>> set(a(1),'Data.VARIA.A1',40); % modify a single Data content
The new assignments apply on the object itself when the method call has no output argument, but can additionally be copied to an other variable:
>> b = set(a(1),'Data.VARIA.A1',40);	% modify the Data content and send back the modified object 
The set assignment may also be used with arrays of iData objects.

This syntax can be simplified by adopting the structure-type one:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a.Data.VARIA.A1
ans =

>> a.Data.VARIA.A1 = 41 % returns the modified object
We shall see below that there is a way to extend an iData object, that is add some new information, and to create links within the object.

The Title of an object usually stores the origin and type of the data (filename).
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a.Title
ans =

UPt3;File sv1850.scn ILL TAS Data

In addition, an object Label can be defined to e.g. sort objects with user defined names, such as 'Measurement1',',Background', ... in the spirit of Labels (Mac OSX) and Emblems (Linux/Gnome) in window managers to provide customized colours and/or icons.

>> a.Label = 'TAS'
a = iData object:

[Tag] [Dimension] [Title] [Last command] [Label]
tp120236 [15 1] 'UPt3;File sv1850.scn ILL TAS...' tp120236=set(tp12023... TAS
These can be searched using the findfield, and findstr methods, see above, or findobj(iData,'Label','TAS').

Definition of Aliases: links and values (setalias, getalias, set, get)

An Alias in iData objects correspond to user fields, dynamically created and evaluated. The previously seen UserData field of the iData structure is static. One can indeed store something there, but there is only one such slot. We shall see that not only we can create new fields to hold some information, but also that these Aliases can point onto other aliases or parts of the iData object.

The main methods used to assign and extract Alias definitions are setalias and getalias respectively. We shall see below that the set and get methods can also be used, but only apply on the Alias values (which is usually a numerical one).

defining a new field as a value

The method used to define (or re-define) an Alias is setalias:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> setalias(a,'NewField',42)
In this latter case, that is when Aliases are defined as numerics (scalar, vector, matrix), this is strictly equivalent to
>> a.NewField = 42
>> set(a,'NewField',42)
The mechanism seen here works for numerical fields, but will fail if you decide to store structures and cell arrays into an Alias. Then, we recommend to define a new member in the Data part of the object, and then point to it as a link, as explained below.

Alias names are case insensitive, that is a.newfield will also work.

In addition, each Alias can be associated with a label/comment when creating it:
>> setalias(a,'NewField',42,'Answer to the Ultimate Question of Life, the Universe, and Everything')
or with the label method:
>> label(a,'NewField','Answer to the Ultimate Question of Life, the Universe, and Everything')
The current label of an Alias is inquired similarly with the label method
>> label(a,'NewField')
ans =

Answer to the Ultimate Question of Life, the Universe, and Everything
The alias labels are used as axis labels when plotting objects.

getting the value of a new field

The method used to extract the definition of an Alias is getalias:
>> getalias(a,'NewField')
ans =

In this case, that is when Aliases are defined as numerics (scalar, vector, matrix), this is strictly equivalent to
>> a.NewField
>> get(a,'NewField')

defining a new field as a link (reference)

The same setalias method can be used to assign an Alias as an expression. In the following example, a QH alias has been automatically defined as a numeric vector when importing the file, and we define a new Alias which points to it:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a.QH
ans =

>> setalias(a,'NewField','QH')
which makes it so that NewField is de-facto QH.
Links which can not be evaluated are returned as is, that is as character strings.
As explained below, it is possible to define links to othet links, or external files.
Links to other links
More complex links may be defined to other parts of the object or to an other Alias,
>> setalias(a,'NewField','Data.ZEROS')		% a link to the whole ZEROS subfield structure
>> setalias(a,'NewField','Data.ZEROS.A1') % a link to the A1 value in the ZEROS subfield
>> setalias(a,'AnotherNewField', 'NewField'); % creates a link to a link to a Data part (follow me !)
>> a.AnotherNewField
ans =

The AnotherNewField points to NewField, which was defined as Data.ZEROS.A1.

This brings the ability to enrich the Data area of the object with new content, and then define an Alias which links to it:
>> a.Data.NewContent1 = rand(10);			% add a new array in the Data
>> a.Data.NewContent2 = struct('a',1,'b', rand(10)); % add a new structure in the Data
>> setalias(a,'HowAreYouToday', 'Data.NewContent2');
Links as more general expressions
Advanced Alias definition may be set by means of the 'this' self-referring symbol which can be used within full expressions. 'this' then refers to the object itself.
>> setalias(a,'NewField', '[ this.Data.ZEROS.A1 this.Data.VARIA.A1 ]') 	% an expression which is built from two links
>> a.NewField

ans =

-115.7500 39.2600
>> a.AnotherNewField

ans =

-115.7500 39.2600
In this last example, we have changed the NewField definition, which immediately affects AnotherNewField (defined above).
The links are evaluated dynamically, so that their value follow the target one.
Links to external resources
In the case the link begins with file://, ftp://, http://, or https:// the external resource is accessed when requesting the alias value, and its full content is retrieved as a structure using iLoad. The external link may optionally end with an anchor specification #keyword, and the matching elements in the external file are then selected and returned.
>> setalias(a,'NewField', [ 'file://' ifitpath '/Data/sv1884.scn#Data' ]) 
The file is accessed for each alias value request.

getting the definition of a new field (reference)

In the previous examples, the QH Alias has been defined from the import routine, and it is assigned a static array (you can see that by looking at the get(a) output).
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> setalias(a,'NewField','QH')
>> get(a,'NewField') % is the same as a.NewField: return the value
ans =

>> getalias(a,'NewField') % return the definition
ans =

We see here that the get method follows definitions recursively until a numeric value is found.
The getalias returns the current Alias definition, but does not propagate to other potential links.

assigning the value of a linked field

When assigning a value to an Alias with the equal sign and the set method, the value is propagated until the fully qualified path is found. This means that when an Alias is a linked field, setting its value will affect the target, but not the Alias itself:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> setalias(a,'NewField',3) % a static scalar
>> setalias(a,'AnotherNewField', 'NewField');
>> a.NewField
ans =

>> a.AnotherNewField = 4 % this stores 4 in the NewField alias !
The direct structure-like assignment and set method propagate the assignment value to the fully qualified link.

getting the value of a linked field

The direct structure-like reference and get method propagate the fully qualified link in order to actually find its value.
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> setalias(a,'NewField',3) % a static scalar
>> setalias(a,'AnotherNewField', 'NewField');
>> get(a,'AnotherNewField') % evaluated as NewField, then as 3
ans =

>> a.AnotherNewField % returns the same as get

removing a field from the object

Added Alias fields can be removed/deleted by setting their value to empty or using the rmalias method
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> setalias(a,'NewField','PNT') % point to PNT, which is a vector here
>> setalias(a,'NewField','') % removes NewField alias
>> rmalias(a,'NewField')
whereas the similar set command removes the final target PNT but leaves the NewField Alias pointing to nothing.
>> setalias(a,'NewField','PNT')
>> set(a,'NewField','') % removes PNT, leaves NewField in error state (undefined)
Re-defining PNT will, at the same time, re-enable NewField.

A warning is in place here

As the set and get methods follow the Aliases definitions until the final value is obtained, this also applies to re-definitions of Aliases. In the previous example, if we now use the set method, we shall modify the final definition of the Alias, that is the one obtained by following all definitions:
>> setalias(a,'NewField','QH')
>> set(a,'NewField',1) % will set QH to 1
That's why we strongly recommend to use setalias when you wish to control an Alias immediate definition, and set when you want to control its final value. The setalias can also define the Alias as a static numerical value, as seen previously, without risking to affect deeper object parts. Similarly, a call to getalias will reveal the nature of the Alias, as link or numerical value.

An other example would be to define QH as a column of a matrix. Changing its value, or the one of any further Alias pointing on it will effectively modify the column in the matrix itself.

In a few words, the set/get functions work just as the linking mechanism in Linux/Unix systems with the ln command.

Definition of the Signal, Error bar and Monitor

The Signal is a special Alias that points to a numerical block which is the relevant data to handle for plotting or computing (see mathematical operations).

As an alias, its value and definition can be accessed with all the following methods (refer to the Alias section above):
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])	% returns an iData object
>> a.Signal % returns the Signal value as an Alias
>> get(a,'Signal')
>> getalias(a,'Signal') % returns the Signal definition, which here points to the CNTS Alias
In the iData terminology, the Signal/Monitor is the rank 0 axis (see the Axis section below)
>> getaxis(a, 0 )	% request Signal/Monitor value as an axis
>> a{0}
Setting the axis 0 with a value first multiplies it by the Monitor, and then sets the Signal. In case you want to ignore the Monitor weighting, simply assign the Signal directly.

The Signal label can be inquired or set with the usual setalias and label methods seen in the Alias section.
>> label(a,'Signal')			% returns the current Signal label
>> label(a,'Signal', 'Make it so ?')
>> label(a, 0, , 'Make it so ?')
This label is shown on plots along the Signal axis (e.g. Y on 1D plots and Z on 2D plots). We point out that the object can be assigned a global Title and Label as well. The title is shown as the plot title and the object text display (see Inquiring the iData object).

It is also possible to extract a portion of the signal from an iData object, similarly as for a normal array:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a([ 2 4 6 8 9])
creates a new iData object with a subscript Signal (and relevant axes definitions - see below).

Dimensionality and Types of Signals that can be handled

The dimensionality of the iData object is the one from its Signal. It is obtained from the ndims method and the length of each dimension is obtained with the size method.
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> ndims(a)
ans =

>> size(a)
ans =

15 1
In principle, there is no limitation concerning the size of the Data and Signal, and the dimensionality (except memory restrictions). The following table indicates the different types of Signal that the iData can handle, and the associated requirements regarding Axes.

Signal size
Related axes
empty ndims=0
Signal is empty
No axis
0D scalar ndims=1
Signal is a single number
In principle, no axis is needed here.
1D vector ndims=1
Signal is a vector
One Axis required, as scalar or vector
2D matrix ndims=2
Signal is a matrix
Two axes required, as scalar, vector or matrix matching Signal
3D volume ndims=3
Signal is a volume
Three Axes are required, as scalar, vector, matrix or volume matching Signal
nD grid space ndims=n
Signal is nD
n Axes are required, each may have a dimensionality from 1 to n, with preference for vectors.
nD event - column based ndims=n
Signal is a long vector (1D)
n Axes are required, each being a scalar or vector of same length as Signal. NOTE: Not all Math methods support this type of data, and it should be converted, when appropriate, into nD grid histogramed with the hist or interp operators (see Math/interpolation).

The first dimension refers to rows, the second to columns, the third to pages,  and so on.

Error bars definition

The Signal is attached to an Error bar, which represents an uncertainty on a Signal value. The Error bar is considered to be symmetric w.r.t. each Signal value. This Error propagates along most mathematical operators and is also used for proper weighting of optimization methods when performing model fit using e.g. a least-square criteria. No error bar mechanism is currently implemented for axes and aliases.
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a.Signal
>> a.Error
>> getalias(a,'Signal')
ans =

>> getalias(a,'Error')
ans =


The Error defaults to sqrt(Signal) when not set (that is a.Error=[]). A constant Error can be set, an expression (including using Data or this) or any numerical value that matches the Signal size.

Monitor definition

Similarly, a Monitor is attached to the Signal. It represents e.g. a counting time or any statistical weight w=1/Monitor that is to be taken into account when performing mathematical operations such as combine and most binary operators (+-/*) involving normalization between signals. In these cases, the Signal and Error are divided by the Monitor prior to performing the operation, and multiplied back after operation.

The Monitor defaults to 1 when not set (that is a.Monitor=[]). A constant Monitor can be set, an expression (including using Data or this) or any numerical value that matches the Signal size.

The Signal, Error and Monitor fields of an iData object are protected Aliases: they can not be deleted.

The raw Signal and Error are obtained with:
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> a.Signal
>> a.Error
>> a=load(iData, [ ifitpath 'Data/sv1850.scn'])
>> getaxis(a, 'Signal') % or: a{0}
>> getaxis(a, 'Error')

Handling event lists

Some data sets contain lists of 'events', that is a set of coordinates for a set of signal values. These events are then collections of 'rows', which each has structure:

[ x y z ... sx, sy, sz ...]

where x,y,z,... are the coordinates, and sx,sy,sz,... are signal values (for instance a magnetic field vector at given location in space).
On import, these data sets result in a matrix which rows are events, and columns are the coordinate/values, as above.

It is possible to format the data set into a formal event list, using the 'event' method after loading. The 'event' method can either make a guess on the dimensionality of space and signal, or specify the number of columns which are affected to the coordinates x,y,z,... (dimensionality of space), and the columns which is to be used for the 'signal'. When more that one 'signal' column is given, the norm of the given columns, per event, is used.

Event data sets can further be converted into histograms using the 'hist' or 'meshgrid' methods. The histograms can be prettified using the 'fill' method, which makes sure that empty spaces in the histogram are interpolated with neighboring values. More information available in the Math page.