Developing an Import Converter
Mathematica provides functions that allow users to write their own file format converters and integrate them with the Mathematica Import and Export framework. You can implement format converters and use Import to import data from arbitrary formats.
Mathematica also includes source code that illustrates how to implement and register format converters. These can be found in the folders $InstallationDirectory/SystemFiles/Formats/format, where format is one of the following: BDF, DIF, MTP, SMILES, SurferGrid, TGF, or TLE. The registration code is placed in the files Import.m or Export.m, and the converter implementations reside in the file Converter.m.
The interface between Import and low-level converter functions is specified by 
(under the 
context). In essence, 
tells the Import and Export framework how to call specific functions when importing specific elements of a file format.
The following terminology is used throughout this tutorial:
A low-level function takes a file or stream as input and returns a list of rules containing the imported data. There are two types of low-level functions: (1) the default importer, which is called by the framework when importing an element not explicitly registered; and (2) the conditional importer, which imports a specific element registered in the second argument of 
.
A post-importer or post-import function, registered in the third argument of 
, takes as input the output of the low-level functions.
There are several forms of 
, summarized below. Throughout this tutorial are progressively advanced examples, showing all uses of 
in detail.
| ImportExport`RegisterImport["format",defaultFunction] |
| register a single defaultFunction to be used by the Import framework as default importer when importing a file of the type  |
| ImportExport`RegisterImport["format",{"elem1"⧴conditionalFunction1,"elem2"⧴conditionalFunction2,...,defaultFunction}] | register multiple elements ( ,  , ...) and respective converter functions ( ,  , ...) to be used by the Import framework; also register defaultFunction to be used when an element requested does not match any registered elements |
| ImportExport`RegisterImport["format",{conditionalFuncs,defaultFunction},{"elem3"⧴postFunction3,"elem4"⧴postFunction4,...}] | register additional converter functions whose input is the output of one of the low-level functions |
Default Importer
For example, suppose you have a file format containing three header lines followed by four columns of numbers.
Registration and Implementation of a New Format
One possible design is to import the header information and the numbers, respectively, to the 
and 
elements. This can also be implemented using 
.
In this particular case, you are telling the Import and Export framework to call the function 
when importing any element of the format 
.
By default, the framework passes the file name to the low-level function, so 
takes as input the file name and a set of options. This function must return a list of rules in the form of 
.
Importing a File of the New Format
Import can now use 
as a valid file format.
| Out[4]= |  |
| Out[5]= |  |
| Out[6]= |  |
Conditional Raw Importers
When a format contains many elements, it may be useful and efficient to import specific elements with specific low-level functions. This can be achieved by giving a list of rules in the form of "elem"->func as the second argument of 
. The list, however, must end with the name of the default importer, which is called when importing elements that do not match any explicitly defined in the list.
Registration and Implementation of a New Format with Conditional Importers
This registration tells the Import and Export framework how to import files of the format 
:
(1) use 
when importing the 
element, and
(2) 
for all other elements.
The low-level functions again have the same structure, taking a file name and (optionally) a list of options, and returning a set of rules in the form of 
.
Import Using MyFormat2
The output of the import elements of 
is the same as those of 
, but now two different functions are called for the two different elements.
| Out[10]= |  |
| Out[11]= |  |
Specifying Subelements
By default, the framework imports subelements using Part.
| Out[12]= |  |
For files containing several large datasets, it may be efficient to directly import specific datasets. For example, you can directly import a dataset from a file with the "EDF" file format.
| Out[13]= |  |
| Out[38]= |  |
You can specify the import of subelements by registering a low-level function in the form of
.
The output of the low-level function must match the form 
.
As before, the output of the other low-level functions must be a list of rules in the form of 
.
The import of string subelements now calls the appropriate low-level function.
| Out[20]= |  |
| Out[21]= |  |
Post-Importers
It may be the case that you have to build elements based on other elements. For example, if the data to be imported is a list of numbers representing a grayscale image, then importing the 
element requires first importing the 
element. In this section are two examples of this using the 
and 
elements.
The post-importer takes as input the output of the conditional importer when a matching element name exists; otherwise, the post-importer takes as input the output of the default importer.
Unlike conditional and default importers, the post-importer simply returns the value of the element.
Registration and Implementation of a New Format with Post-Importers
To illustrate the differences between a conditional importer and a post-importer, 
is extended with two additional elements: 
and 
. The 
element is imported via a conditional importer. The 
element, however, is imported via a post-importer.
The registration below tells the Import and Export framework how to import files of the format 
:
(1) for the 
or 
elements, call the corresponding conditional importers,
(2) for the 
element, call the default importer first, and use its output as input for 
, and
(3) for all other elements, call the default importer.
The conditional and default importers have the same structure as before.
Notice that the 
importer has to explicitly call the default importer and extract the data manually.
Since no 
element is registered as a conditional importer, the importer of the 
element takes as input the output of the default importer.
Import Using MyFormat3
From a user perspective, there is no difference between an element implemented using a post-import function or low-level function.
The
element is registered as a conditional importer.
| Out[27]= |  |
The import of the
element calls the post-importer
.
| Out[28]= |  |
Options to RegisterImport
has several options that allow great flexibility.
"FunctionChannels" and "BinaryFormat"
In our example above, the low-level functions accept a file name as an argument, and the functions open a stream to the file. The framework can directly pass an InputStream to the low-level functions by specifying 
as an option to 
.
By specifying the option "BinaryFormat"->True, the framework passes a binary stream to the low-level importer.
The default value of 
is 
. The default value of 
is False.
Example
For the format registered as
the signature of eFunc is 
, and the framework passes a (non-binary) stream to eFunc.
"AvailableElements"
By default, when importing an element not explicitly registered as a conditional importer or a post-importer, the framework evaluates the default importer. If no matching element is found in the default importer, the framework generates an error message and returns $Failed.
By specifying the option 
when attempting to import an element not present in the specified list, the framework will directly return $Failed and generate an error message without calling any low-level importer.
Example
For the format registered as
when you call Import[filename, {"format", "foo"}], the framework will return $Failed without evaluating the default importer eDefaultFunc.
Note that it is an error to specify 
. In this case, Import[filename, {"format", "elem2"}] will return $Failed because 
is not in the list specified by 
.
"DefaultElement"
Specifying "DefaultElement"->elem, where elem is the name of an element, the framework imports the elem when no Import element is specified.
"Sources"
The option 
can be used to specify file paths to .m, .mx, or MathLink .exe files that contain definitions of the low-level functions. The framework will automatically use Get or Install appropriately for the source files.
