This tutorial is designed for new HDF5 users. We will go over a brief history of HDF and HDF5 software, and will cover basic HDF5 Data Model objects and their properties; we will give an overview of the HDF5 Libraries and APIs, and discuss the HDF5 programming model. Simple C and Fortran examples, and Java tool HDFView will be used to illustrate HDF5 concepts.
1. Introduction to HDF5
HDF & HDF-EOS Workshop XII
October 15, 2008
10/15/08
HDF & HDF-EOS Workshop XII 1
1
2. Topics Covered
- Introduce HDF5
- Describe HDF5 Data and Programming Models
- Walk Through Example Code
10/15/08
HDF & HDF-EOS Workshop XII 2
2
3. For More Information …
All workshop slides will be available from:
http://hdfeos.org/workshops/ws12/workshop_twelve.php
10/15/08
HDF & HDF-EOS Workshop XII 3
4. What is HDF5?
HDF = Hierarchical Data Format
• Data model, library and file format for managing
data
• Tools for accessing data in the HDF5 format
10/15/08
HDF & HDF-EOS Workshop XII 4
5. Brief History of HDF
1987 At NCSA (University of Illinois), a task force formed to create an
architecture-independent format and library:
AEHOO (All Encompassing Hierarchical Object Oriented format)
Became HDF
Early NASA adopted HDF for Earth Observing System project
1990’s
1996
DOE’s ASC (Advanced Simulation and Computing) Project began
collaborating with the HDF group (NCSA) to create “Big HDF”
(Increase in computing power of DOE systems at LLNL, LANL and
Sandia National labs, required bigger, more complex data files).
“Big HDF” became HDF5.
1998
HDF5 was released with support from National Labs, NASA, NCSA
2006 The HDF Group spun off from University of Illinois as non-profit
corporation
10/15/08
HDF & HDF-EOS Workshop XII 5
6. Why HDF5?
In one sentence ...
10/15/08
HDF & HDF-EOS Workshop XII 6
6
7. Answering big questions …
Matter and the universe
Life and nature
August 24, 2001
August 24, 2002
Total Column Ozone (Dobson)
60
385
610
Weather and climate
10/15/08
HDF & HDF-EOS Workshop XII 7
7
8. … involves big data …
10/15/08
HDF & HDF-EOS Workshop XII 8
8
9. … varied data …
LCI Tutorial
10/15/08
Thanks to Mark
HDF & HDF-EOS Workshop XII 9 Miller, LLNL
9
10. … and complex relationships …
SNP Score
Contig Summaries
Discrepancies
Contig Qualities
Coverage Depth
Trace
Reads
Aligned bases
Read
quality
Contig
Percent match
10/15/08
HDF & HDF-EOS Workshop XII 10
10
11. … on big computers …
… and small computers …
10/15/08
HDF & HDF-EOS Workshop XII 11
11
12. How do we…
• Describe our data?
• Read it? Store it? Find it? Share it? Mine it?
• Move it into, out of, and between computers and
repositories?
• Achieve storage and I/O efficiency?
• Give applications and tools easy access our data?
10/15/08
HDF & HDF-EOS Workshop XII 12
12
13. Solution: HDF5!
• Can store all kinds of data in a variety of ways
• Runs on most systems
• Lots of tools to access data
• Emphasis on standards (HDF-EOS, CGNS)
• Library and format emphasis on I/O efficiency and
storage
10/15/08
HDF & HDF-EOS Workshop XII 13
14. Structure of HDF5 Library
Applications
Object API (C, F90, C++, Java)
Library internals
Virtual file I/O
File or other “storage”
10/15/08
HDF & HDF-EOS Workshop XII 14
16. HDF5 Applications & Domains
Examples: Thermonuclear simulations
Product modeling
Data mining tools
Visualization tools
Climate models
Simulation, visualization,
remote sensing…
HDF-EOS
Virtual File Layer
(I/O Drivers)
Stdio
CGNS
HDF5 Data Model & API
Split Files
MPI I/O
Storage
HDF5
format
10/15/08
File
ASC
Custom
?
Split metadata File on parallel
and raw data files file system
User-defined
device
HDF & HDF-EOS Workshop XII 16
Communities
17. Lots of Layers in HDF5!
“Ogres are like onions.”
Shrek HDF5 Monster??
Just like Shrek, once you get to
know HDF5 you will really like it!!
10/15/08
HDF & HDF-EOS Workshop XII 17
19. An HDF5 file is a container…
…into
which you
can put
your data
objects.
10/15/08
lat | lon | temp
----|-----|----12 | 23 | 3.1
15 | 24 | 4.2
17 | 21 | 3.6
te
let
pa
HDF & HDF-EOS Workshop XII 19
19
21. HDF5 Data Model
Primary Objects
• Groups
• Datasets
Additional ways to organize and annotate data
• Attributes
• Storage and access properties
Everything else is built from these parts.
10/15/08
HDF & HDF-EOS Workshop XII 21
21
23. Dataspaces
Two roles:
• Dataspace contains spatial info about a dataset
stored in a file
• Rank and dimensions
• Permanent part of dataset
definition
Rank = 2
Dimensions = 4x6
• Partial I/0: Dataspace describes application’s data
buffer and data elements participating in I/O
Rank = 1
Dimension = 10
10/15/08
HDF & HDF-EOS Workshop XII 23
23
24. Write – from memory to disk
memory
10/15/08
disk
HDF & HDF-EOS Workshop XII 24
24
25. Partial I/O
Move just part of a dataset
memory
disk
(a) Slab from a 2D array to the
corner of a smaller 2D array
Elements in each must be same.
10/15/08
disk
memory
(b) Regular series of blocks from a
2D array to a contiguous sequence
at a certain offset in a 1D array
HDF & HDF-EOS Workshop XII 25
25
26. Datatypes (array elements)
• Datatype – how to interpret a data element
• Permanent part of the dataset definition
• Two classes: atomic and compound
10/15/08
HDF & HDF-EOS Workshop XII 26
26
27. Datatypes
• HDF5 atomic types include:
integer & float
user-definable (e.g., 13-bit integer)
variable length types (e.g., strings)
references to objects/dataset regions
enumeration - names mapped to integers
• HDF5 compound types
Comparable to C structs (“records”)
Members can be atomic or compound types
10/15/08
HDF & HDF-EOS Workshop XII 27
27
28. HDF5 dataset: array of records
3
5
Dimensionality: 5 x 3
int8
int4
int16 2x3x2 array of float32
Datatype:
Record
10/15/08
HDF & HDF-EOS Workshop XII 28
28
29. Properties
• Properties are characteristics of HDF5 objects
that can be modified
• Default properties handle most needs
• By changing properties can take advantage of the
more powerful features in HDF5
10/15/08
HDF & HDF-EOS Workshop XII 29
30. Special Storage Properties
Better subsetting
access time;
extensible
chunked
Improves storage
efficiency,
transmission speed
compressed
Arrays can be
extended in any
direction
extensible
File B
split file
Dataset “Fred”
File A
Metadata for Fred
10/15/08
Metadata in one file,
raw data in another
Data for Fred
HDF & HDF-EOS Workshop XII 30
30
31. Attributes (optional)
• Attribute – data of the form “name = value”,
attached to an object
• Operations similar to dataset operations, but …
Not extensible
No compression or partial I/O
• Can be overwritten, deleted, added during the
“life” of a dataset
10/15/08
HDF & HDF-EOS Workshop XII 31
31
33. Groups
• A mechanism for organizing collections
• Every file starts with a root group
• Similar to UNIX directories
A
• Can have attributes
k
10/15/08
“/”
B
l m
HDF & HDF-EOS Workshop XII 33
33
C
34. Path to HDF5 Object in a File
/ (root)
/x
/foo
/foo/temp
/foo/bar/temp
10/15/08
foo
temp
“/”
x
bar
temp
HDF & HDF-EOS Workshop XII 34
34
37. Useful Tools For New Users
h5dump:
Tool to “dump” or display contents of HDF5 files
h5cc, h5c++, h5fc:
Scripts to compile applications
HDFView:
Java browser to view HDF4 and HDF5 files
10/15/08
HDF & HDF-EOS Workshop XII 37
38. H5dump Command-line Utility To View HDF5 File
h5dump [--header] [-a ] [-d <names>] [-g <names>]
[-l <names>] [-t <names>] [-p] <file>
--header
Display header only; no data is displayed.
-a <names> Display the specified attribute(s).
-d <names> Display the specified dataset(s).
-g <names> Display the specified group(s) and all the members.
-l <names>
Displays the value(s) of the specified soft link(s).
-t <names> Display the specified named datatype(s).
-p
Display properties.
<names> is one or more appropriate object names.
10/15/08
HDF & HDF-EOS Workshop XII 38
46. Simple HDF5 File in HDFView
Right-click and select
“Open” with mouse
Right-click and select
“Show Properties”
with mouse
10/15/08
HDF & HDF-EOS Workshop XII 46
47. Simple HDF5 File in HDFView
10/15/08
HDF & HDF-EOS Workshop XII 47
52. Operations Supported by the API
• Create objects (groups, datasets, attributes, complex data
types, …)
• Assign storage and I/O properties to objects
• Perform complex subsetting during read/write
• Use variety of I/O “devices” (parallel, remote, etc.)
• Transform data during I/O
• Make inquiries on file and object structure, content,
properties
10/15/08
HDF & HDF-EOS Workshop XII 52
52
53. General Programming Paradigm
• Properties of object are optionally defined
Creation properties
Access property lists
• Object is opened or created
• Object is accessed, possibly many times
• Object is closed
10/15/08
HDF & HDF-EOS Workshop XII 53
53
54. Order of Operations
• An order is imposed on operations by argument
dependencies
For Example:
A file must be opened before a dataset
-becausethe dataset open call requires a file handle
as an argument.
• Objects can be closed in any order.
10/15/08
HDF & HDF-EOS Workshop XII 54
54
55. The General HDF5 API
• Currently C, Fortran 90, Java, and C++ bindings.
• C routines begin with prefix H5?
? is a character corresponding to the type of object
the function acts on
Example Functions:
H5D : Dataset interface
H5F : File interface
e.g., H5Dread
e.g., H5Fopen
H5S : dataSpace interface e.g., H5Sclose
10/15/08
HDF & HDF-EOS Workshop XII 55
55
56. HDF5 Defined Types
For portability, the HDF5 library has its own defined
types:
hid_t:
hsize_t:
hssize_t:
object identifiers (native integer)
size used for dimensions (unsigned long or
unsigned long long)
for specifying coordinates and sometimes for
dimensions (signed long or signed long long)
herr_t:
function return value
hvl_t:
variable length datatype
For C, include hdf5.h in your HDF5 application.
10/15/08
HDF & HDF-EOS Workshop XII 56
56
57. The HDF5 API
• For flexibility, the API is extensive
300+ functions
Victronix
Swiss Army
Cybertool 34
• This can be daunting… but there is hope
A few functions can do a lot
Start simple
Build up knowledge as more features are needed
10/15/08
HDF & HDF-EOS Workshop XII 57
57
58. Basic Functions
H5Fcreate (H5Fopen)
H5Screate_simple
H5Dcreate (H5Dopen)
H5Dread, H5Dwrite
H5Dclose
H5Sclose
H5Fclose
10/15/08
create (open) File
create dataSpace
create (open) Dataset
access Dataset
close Dataset
close dataSpace
close File
HDF & HDF-EOS Workshop XII 58
60. High Level APIs
• Included along with the HDF5 library
• Simplify steps for creating, writing, and reading
objects
• Do not entirely ‘wrap’ HDF5 library
10/15/08
HDF & HDF-EOS Workshop XII 60
62. Steps to Create a File
1. Decide on special properties the file should have
•
•
•
Creation properties, like size of user block
Access properties, such as metadata cache size
Use default properties (H5P_DEFAULT)
2. Create property lists, if necessary
3. Create the file
4. Close the file and the property lists, as needed
10/15/08
HDF & HDF-EOS Workshop XII 62
62
63. Code: Create a File
hid_t
herr_t
file_id;
status;
file_id = H5Fcreate ("file.h5", H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT);
status = H5Fclose (file_id);
“/” (root)
Note: Return codes not checked for errors in code samples.
10/15/08
HDF & HDF-EOS Workshop XII 63
63
65. Steps to Create a Dataset
1. Define dataset characteristics
•
•
•
Dataspace - 4x6
Datatype – integer
Properties if needed, or use H5P_DEFAULT
2. Decide where to put it
•
Obtain location ID:
- Group ID puts it in a Group
- File ID puts it in Root Group
“/” (root)
A
3. Create dataset in file
4. Close everything
10/15/08
HDF & HDF-EOS Workshop XII 65
65
66. HDF5 Pre-defined Datatype Identifiers
HDF5 defines* set of Datatype Identifiers per HDF5
session.
For example:
C Type
HDF5 File Type
HDF5 Memory Type
int
H5T_STD_I32BE
H5T_STD_I32LE
H5T_NATIVE_INT
float
H5T_IEEE_F32BE
H5T_IEEE_F32LE
H5T_NATIVE_FLOAT
double
H5T_IEEE_F64BE
H5T_IEEE_F64LE
H5T_NATIVE_DOUBLE
* Value of datatype is NOT fixed
10/15/08
HDF & HDF-EOS Workshop XII 66
67. Pre-defined File Datatype Identifiers
Examples:
H5T_IEEE_F64LE Eight-byte, little-endian, IEEE floating-point
H5T_STD_I32LE Four-byte, little-endian, signed two's
complement integer
Architecture*
Programming
Type
NOTE: What you see in the file. Name is the same everywhere and
explicitly defines a datatype.
*STD= “An architecture with a semi-standard type like 2’s complement integer, unsigned integer…”
10/15/08
HDF & HDF-EOS Workshop XII 67
68. Pre-defined Native Datatypes
Examples of predefined native types in C:
H5T_NATIVE_INT
H5T_NATIVE_FLOAT
H5T_NATIVE_UINT
H5T_NATIVE_LONG
H5T_NATIVE_CHAR
(int)
(float )
(unsigned int)
(long )
(char )
NOTE: Memory types.
Different for each machine.
Used for reading/writing.
10/15/08
HDF & HDF-EOS Workshop XII 68
69. Dataset Creation Property List
Dataset creation property list: information on how to
organize data in storage.
Chunked
Chunked &
compressed
H5P_DEFAULT: contiguous
10/15/08
HDF & HDF-EOS Workshop XII 69
69
70. Code: Create a Dataset
1
2
3
hid_t
hsize_t
herr_t
file_id, dataset_id, dataspace_id;
dims[2];
status;
4
file_id = H5Fcreate (”file.h5", H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT);
Create a
5 dataspace= 4;
dims[0]
6
7
rank
dims[1] = 6;
dataspace_id = H5Screate_simple (2, dims, NULL);
Create a dataset
8
current dims
pathname
datatype
dataset_id = H5Dcreate(file_id,”A",H5T_STD_I32BE,
dataspace_id, H5P_DEFAULT);
dataspace
Terminate access to dataset, dataspace,
file
9 status = H5Dclose (dataset_id);
10 status = H5Sclose (dataspace_id);
11 status = H5Fclose (file_id);
10/15/08
property list
(default)
HDF & HDF-EOS Workshop XII 70
70
71. Example Code - H5Dwrite
Dataset Identifier from
H5Dcreate or H5Dopen
Memory Datatype
status = H5Dwrite (dataset_id, H5T_NATIVE_INT, H5S_ALL,
H5S_ALL, H5P_DEFAULT, dset_data);
10/15/08
HDF & HDF-EOS Workshop XII 71
72. Example Code – H5Dwrite
status = H5Dwrite (dataset_id, H5T_NATIVE_INT, H5S_ALL, H5S_ALL,
H5P_DEFAULT, dset_data);
Data Transfer Property List
(MPI I/O, Transformations, …)
Memory
Dataspace
File
Dataspace
H5S_ALL selects entire
dataspace
10/15/08
HDF & HDF-EOS Workshop XII 72
73. Partial I/O
Memory Dataspace
H5S_ALL
File Dataspace (disk)
H5S_ALL
Get a Dataspace:
H5Screate_simple
H5Dget_space
Modify Dataspace:
H5Sselect_hyperslab
H5Sselect_elements
10/15/08
HDF & HDF-EOS Workshop XII 73
74. Example Code – H5Dread
status = H5Dread (dataset_id, H5T_NATIVE_INT,
H5S_ALL, H5S_ALL, H5P_DEFAULT, dset_rdata);
10/15/08
HDF & HDF-EOS Workshop XII 74
75. High Level APIs: HDF5 Lite (H5LT)
#include "H5LT.h"
…
file_id = H5Fcreate (“file.h5", H5F_ACC_TRUNC,
H5P_DEFAULT, H5P_DEFAULT);
status = H5LTmake_dataset (file_id,“A", 2, dims,
H5T_STD_I32BE, data);
status = H5Fclose (file_id);
10/15/08
HDF & HDF-EOS Workshop XII 75
77. Example: Create a Group
“/” (root)
A
B
4x6 array of
integers
file.h5
10/15/08
HDF & HDF-EOS Workshop XII 77
77
78. Steps to Create a Group
1. Decide where to put it – “root group”
•
Obtain location ID
2. Decide name – “B”
3. Create group in file
4. (Eventually) close the group.
10/15/08
HDF & HDF-EOS Workshop XII 78
78
79. Code: Create a Group
hid_t file_id, group_id;
...
/* Open “file.h5” */
file_id = H5Fopen (“file.h5”, H5F_ACC_RDWR,
H5P_DEFAULT);
/* Create group "/B" in file. */
group_id = H5Gcreate (file_id,"B",0);
Size hint for number of
bytes to store names of
objects. 0=default
/* Close group and file. */
status = H5Gclose (group_id);
status = H5Fclose (file_id);
10/15/08
HDF & HDF-EOS Workshop XII 79
79
80. Thank you!
This work was supported by the Cooperative Agreement with the
National Aeronautics and Space Administration (NASA) under NASA
grant NNX06AC83A and NNX08A077A. Any opinions, findings,
conclusions or recommendations expressed in this material are those of
the author(s) and do not necessarily reflect the views of NASA.
10/15/08
HDF & HDF-EOS Workshop XII 80
Notas do Editor
The CFD General Notation System (CGNS) provides a general, portable, and extensible standard for the storage and retrieval of computational fluid dynamics (CFD) analysis data. It consists of a collection of conventions, and free and open software implementing those conventions. It is self-descriptive, machine-independent, well-documented, and administered by an international steering committee.
The CGNS implementation of SIDS, so-called MLL, was originally built using a file format called ADF (Advanced Data Format). This format was based on a common file format system previously in use at McDonnell Douglas. The ADF has worked extremely well, requiring little repair, upgrade, or maintenance over the last decade. However, ADF does not have parallel I/O or data compression capabilities, and does not have the support and tools that the storage format HDF5 offers. HDF5, supported by The HDF Group, has rapidly grown to become a world-wide format standard for storing scientific data. HDF5 has parallel capability as well as a broader support base than ADF.
This shows that you can mix objects of different types according to your needs. Typically, there will be metadata stored with objects to indicate what type of object they are.
Like HDF4, HDF5 has a grouping structure. The main difference is that every HDF5 file starts with a root group, whereas HDF4 doesn’t need any groups at all.
Data Array is an ordered collection of identically typed data items distinguished by their indices
Metadata:
Dataspace – Rank, dimensions; spatial info about dataset
Datatype – Information on how to interpret your data
Storage Properties – How array is organized
Attributes – User-defined metadata (optional)
Here is an example of a basic HDF5 object.
Notice that each element in the 3D array is a record with four values in it.
Data Array is an ordered collection of identically typed data items distinguished by their indices
Metadata:
Dataspace – Rank, dimensions; spatial info about dataset
Datatype – Information on how to interpret your data
Storage Properties – How array is organized
Attributes – User-defined metadata (optional)
Like HDF4, HDF5 has a grouping structure.
The main difference is that every HDF5 file starts with a root group, whereas HDF4 doesn’t need any groups at all.
To create this file, we would start by creating the file itself. When you create a file, the root group gets created with it. So every file has at least that one group.