Learning HDF5 with HDFView

This tutorial enables you to get a feel for HDF5 by using the HDFView browser. It does NOT require any programming experience.

Installation

• Download and install HDFView. It can be downloaded from the Download HDFView page.

• Obtain the storm1.txt text file, used in the tutorial.

Begin Tutorial

Once you have HDFView installed, bring it up and you are ready to begin the tutorial.

Topics Covered

Following are the topics covered in the tutorial. The first topic creates the file that is used in the subsequent topics. 

1. Creating a New HDF5 file with a Contiguous Dataset
2. Displaying a Dataset as an Image
3. Creating Attributes
4. Creating a Compressed and Chunked Dataset
5. Creating an Image and a Subset
6. Creating a Table (Compound Dataset)

Topics

Creating a New HDF5 File with a Contiguous Dataset

The steps below describe how to create a file (storm.h5), group (/Data), and a contiguous dataset (/Data/Storm) using HDFView. A group is an HDF5 object that allows objects to be collected together. A dataset is an array of data values. A contiguous dataset is one that is stored as a single block in the HDF5 file.

• Select the File pull-down menu at the top left, and then select New -> HDF5.

• Specify a location and type in storm.h5 for the name of your file, and click on the Save button. You will see the storm.h5 file in the TableView:
  
  

• Right click on storm.h5, and select New -> Group .

• Enter Data for the name of the group and then click the Ok button. You will see the group Data in the TableView.
   

  

• Right click on the group Data and select New -> Dataset.

• A window pops up on the right. Fill in the information as follows, and then click Ok (leave the Datatype information as is):

   

  Dataset Name	Storm
  Under Dataspace, Current size	57x57
  Under Dataspace, Maximum Size	57x57
  Layout	Contiguous (default)

• Left click on the Data group in the tree view to see the Storm dataset in the TableView:

  

• Double left click on the Storm dataset in the tree view. A window with an empty spreadsheet pops open.

• Copy the data from the storm1.txt file into the dataset.
  
  If you downloaded storm1.txt, then right click on the Table menu and select Import Data from Text File. Specify a location, select storm1.txt and click on the Open button. Answer Yes in the dialog box that pops up (which asks if you wish to paste the selected data).

  Alternately, you can copy/paste directly. Select and copy the data in a separate window. Position your cursor at (0,0) in your table, and select Paste from the Table menu.

  The values will be entered into the spreadsheet.
   

  

• Close the dataset, and save the data.

Displaying a Dataset as an Image

Any dataset can be viewed as an image in HDFView. Below are the steps that demonstrate this.

• Right click on Storm in the tree view, and select Open As.

• Select the Image button under Display As (near the top) in the Dataset Selection window that pops up. Then click OK at the bottom of the window to display the image.

  

• The rainbow icon brings you to the Image Palette window. Click on that to play with the palette (GrayWave probably is the best choice). Close.

Creating Attributes

Additional information to describe an object can be stored in attributes. An attribute can be added to a group or dataset with HDFView.

The following illustrates how to add an attribute to the group /Data:

• Click on the /Data folder in the tree view. You will see two tabs, Object Attribute Info and General Object Info, in the pane on the right site of the HDFView window.
  

• With the left mouse button, select the Add Attribute button.

• Select the Add Attribute button to add an attribute with these values:

   

  Name	BatchID
  Type	INTEGER
  Size (bits)	32
  Value	3343

• Select the Ok button. The attribute will show up in the Properties window.

• Close the Properties window.

Adding an attribute to a dataset is very similar to adding an attribute to a group. For example, the following adds an attribute to the /Storm dataset:

• Left mouse click on the /Storm dataset in the tree view. You will see the Object Attribute Info and General Object Info tabs on the right.

• In the Object Attribute Info pane select the Add Attribute button and enter an attribute with these values. (Be sure to add a String Length or the string will bet truncated to one character!):

   

  Name	Units
  Type	STRING
  String Length	3
  Value	m/s

• Select the Ok button. The attribute will be displayed in the window.

• Close the Properties window.

   

Creating a Compressed and Chunked Dataset

A chunked and compressed dataset can be created using HDFView. A compressed dataset is a dataset whose size has been compressed to take up less space. In order to compress an HDF5 dataset, the dataset must be stored with a chunked dataset layout (as multiple chunks that are stored separately in the file).

Please note that the chunk sizes used in this topic are for demonstration purposes only. For information on chunking and specifying an appropriate chunk size, see the Chunking in HDF5 documentation.

Also see the HDF5 Tutorial topic on Creating a Compressed Dataset.

• Right click on storm.h5. Select New -> Group.

• Enter Image for the name of the group, and click the OK button to create the group.
  

   

• Right click on the Image group, and select New -> Dataset.

• Enter the following information for the dataset. Leave the Datatype as is (INTEGER):

   

  Dataset Name	Another Storm
  Under Dataspace, Current size	57x57
  Under Dataspace, Set Max Size	57x57
  Storage Layout	Chunked
  Chunk Size	20x20
  Compression	gzip
  Compression Level	9

      

  You will see the Another Storm dataset in the Image group:
  

  

• Double left-mouse click on the Another Storm dataset to display the spredsheet:

  

• Copy the data from the storm1.txt file into the dataset. (See the previous topic for copying storm1.txt into a dataset.)
• Close the table, and save the data.
• Right click on Another Storm, and select Open As.

• Select the Image button in the Dataset Selection window that pops up. Click the Ok button at the bottom of the window to view the dataset as an image.

  

Creating an Image and a Subset

A previous topic demonstrated how to view any dataset as an image in HDFView. With HDFView you can also create an image to begin with, as is shown below.

• Right click on the Data group and select New -> Image.
• A window pops up on the right. Enter the following and then click Ok:

   

  Image name	Storm Image
  Height	57
  Width	57

• Close the dataset.

• Double left-mouse click on the Data group to see its contents. You will see the Storm Image dataset.

  

• Add data to the Storm Image dataset as was shown previously:
  
  
  • Right click on Storm Image, and select Open As to open the Dataset Selection window.
  • Click on the Spreadsheet button at the top left of the Dataset Selection window to view the image as a spreadsheet.
  • Copy the data from the storm1.txt file into the dataset.
  • Close the dataset and save the data.
• Left double click on Storm Image to see the image. Close the dataset.

• Right click on Storm Image and select Show Properties from the pop-up menu, to open the Properties window. Select the Attributes tab to see the attributes:

  

• Close the Properties window.
• Right click on Storm Image and select Open As to bring up the Data Selection window.

• Select a subset by clicking the left mouse on the image in the window and dragging the mouse. Notice that the Height and Width values change. Select to display it as an image. Click Ok.

  

• Position the cursor in the middle of the image. Press Shift+Left Mouse button and hold, and then drag the mouse to select another subset.
• Select Image->Write Selection to Image. Enter Subset for the new image name. Click Ok. The Subset image will appear in the tree view on the left.

• Left double click on the image Subset to bring it up on the right.

  

• Close the Subset image.

Creating a Table (Compound Dataset)

A dataset with a compound datatype contains data elements that consist of multiple fields. If the dataspace for the compound dataset is one-dimensional, then the dataset can be viewed as a table in HDFView, as is shown below.

• Right button click on the group Data. Select New -> Compound DS.

• A window pops up on the right side of the screen. Only fill in the following fields:

  Dataset name	Table
  Dataspace (Current size only)	4
  Compound Datatype Properties:     Number of Members	3
  Compound Datatype Properties:   Name / Datatype / Size	Description / string / 4 Temperature / float / 1 Pressure / double / 1

    

  
  

  

  Click Ok at the bottom.

• Open the Data group (if it is not open) and double left click on the Table object.

  
  

  

  Close the dataset.