hdf5 abstract data model

02/21/14 1

HDF5 Abstract Data Model

Classes, Objects and their relationships

http://hdf.ncsa.uiuc.edu/HDF5/ADM_EOS_Sep99/index.htm

02/21/14 2

Outline

• Motivation and goals

• Overview of HDF5 file structure and HDF5 objects

• UML notation

• Formal description of HDF5 file, HDF5 objects and operations in UML

02/21/14 3

Motivation and Goals

• Need a formal description of HDF5 file and library• Need an HDF5 abstract data model (ADM) for high level

C++, F90, Java, Ada(?) APIs• Attempt to look at HDF5 ADM from OO point of view • HDF group does not have experience in OO design and

languages

• Universal Modeling Language (UML) has been chosen to document current HDF5 ADM and HDF5 Library

• Suggestions and help from HDF5 “OO users” is WELCOME!

02/21/14 4

• Container for storing scientific data– Primary Objects:

– Groups– Datasets

– Secondary Objects:– Datatypes– Dataspaces

• Additional means to organize data– Attributes– Sharable objects– Storage and access properties

HDF5 File (conceptual view)

02/21/14 5

Example of an HDF5 File

File

GroupDatatype Dataset

Group Dataset

RGroup/

A B C

D E

F

GH

K

Dataset path : /B /C/D/G

02/21/14 6

Example of an HDF5 File (cont.)

File


Group

RGroup/

B C

D

Dataset

E

F

G

A

H

K

Dataset path : /B /C/D/G

02/21/14 7

Example of an HDF5 File (cont.)

File RGroup


Group

/

B C

D

F

K

Dataset path : /B

02/21/14 8

HDF5 File (logical view)

• Directed graph with an entry point (root group)• nodes are HDF5 objects :

– Group, Dataset, Dataspace, Datatype

– each object may have more than one path to it

• edges are inclusion directions (HDF5 links)• graph may have:

– loops– isolated nodes – “dangling” edges

02/21/14 9

UML Notation

Book:

UML Distilled Applying the Standard Object Modeling Language

by Martin Fowler Booch Jacobson Rumbaugh Object Technology Series

Addison-Wesley

02/21/14 10

UML Notation

Class

Class Name Class Name____________ Attribute____________

Operation

02/21/14 11

UML Notation

Association Describes connection between object instances, should be a verb.

Class A Class B

Multiplicity of Association Object A has zero or more objects B associated with it. Object B belongs to at least one and up to N objects A

1..N 0...*

Object A has object B associated with it.

Class A Class B

02/21/14 12

UML Notation

Class A

Generalization(Inheritance) “is a” relationship B is an A

Class B

02/21/14 13

UML Notation

Class A

Aggregation “a part of” relationship Objects B and C are part of object A

Class B Class C

02/21/14 14

UML Notation

Class A

Composition “a part of” relationship Objects B and C “live and die” with A

Class B Class C

02/21/14 15

HDF5 Classes

– File Class– Group Class– Dataset Class– Dataspace Class

– Datatype Class

– Named_Object Class• Group

• Dataset• Named Dataspace• Named Datatype

– User_Defined_Attribute Class

02/21/14 16

HDF5 Classes

– Link Class• Root Link • Non-Root Link

– Soft Link

– Hard Link

– Property List Class• File Property List

– Creation Property List

– Access Property List• Dataset Property List

– Transfer Property List– Storage property List

02/21/14 17

File Class File , Group Class and Root Link Class Association Diagram

1 1File Class Group Class

File is a composition of a Group.Root Link is created when File is created.Root Link points to the Group which is called Root Group.Root Group is automatically created/opened/closed when File is

created/opened/closed.

Root Link Class1

1

File Property List Class1

0…*

02/21/14 18

File Class

______________________________________________

_

open/closecreateis_hdf5get_create_prpget_access_prpmount/unmountreopen

File and Group Classes

Group Class__________________________________________

open/closeiterateget_object_infoget_link_valueget/set_comment

02/21/14 19

Group Class

Non-Root Link Class

0…*

1

Group Class and Associated Classes

Named Object Class

0…1

1…*

Group

HDF5 object HDF5 object...

A B

C

D

02/21/14 20

Non-Root Link Class

Link Class

Soft Link Hard Link

Named Object Named Object10...1

0...* 1...*

Root Link Class

Link Class

02/21/14 21

Non-Root Link Class____________________________

name______________________link/unlinkcreatemove (rename)

Non-Root Link Class

02/21/14 22

Named_Object Class

Group Class Dataset Class Named Datatype Class Named Dataspace Class

Named_Object Class

Soft Link Class User Defined Attribute 0…* 0…*

1 0...1

Hard Link Class 1...*

1

Not implemented yet

02/21/14 23

Dataset

Storage Property List

User Defined Attribute

0…*

0...*

1

1

0...*

1...*

Hard Link Soft Link

1 Dataspace1

Datatype1

0...* 0...*

Dataset Class and associated Classes

1

02/21/14 24

Dataset Class

Dataset Class________________________

user defined attribute________________________

createopen/closewrite/readextendget_space/type/property

02/21/14 25

Attribute Class

Datatype Class Dataspace Class

Named Object Class

1

1 1

Attribute Class and associated Classes

0...*

0...* 0...*

02/21/14 26

Attribute Class

Attribute Class_________________________

name_____________________createopen/closewrite/readget_space/type/property

02/21/14 27

Property List Class and Subclasses

Property List

File Property List Dataset Property List

Creation PList Access PList Storage PList Transfer PList

02/21/14 28

Dataspace Class

Dataspace Class

Hard Link10…*

Transient Dataspace Class

02/21/14 29

Dataspace ClassDataspace Class_______________________________rankcurrent_sizemaximum_size_______________________________create/create_simpleopen/close/lockis_simplecopyextent_classextent_copyset_extent_simpleset_extent_noneget_simple_extent_npoints/ndims/dims

Extent operations

General operations

Dataspace Extent

02/21/14 30

Transient Dataspace Class

Dataspace Class______________selection______________select_elementsselect_noneselect_hyperslabselect_validoffset_simple………………...

Set and modify selections

Dataspace Selection

02/21/14 31

Datatype Class

Datatype Class and Subclasses

Atomic

Compound

Variable Length

Enumeration

Array

02/21/14 32

Datatype Class____________________________copyequalcloseconvertget_classget_size

Datatype Class

02/21/14 33

Atomic Datatype Class and Subclasses

Atomic Datatype Class

Predefined Atomic Userdefined AtomicH5Tcopy

Link Class

0…*

1

02/21/14 34

Predefined Atomic Datatype Class

Predefined Atomic Datatype Class

Reference

Object Reference Region Reference

Opaque

Time String Integer Float

Bitfield

02/21/14 35

Examples of Predefined Atomic Datatype

Predefined Atomic Datatype_______________________name = H5T_arch_base_______________________

H5T_IEEE_F64LE Eight-byte little-endian, IEEE floating-point

H5T_STD_U16BE Two-byte big-endian, unsigned integer

H5T_C_S1 One-byte null-terminated string of eight-bit

characters

H5T_CRAY_F64 Eight-byte Cray floating point

H5T_STD_ROBJ Reference to entire object in a file

H5T_NATIVE_LONG long

02/21/14 36

Compound Datatype Class

Compound Datatype Class

Datatype Class

1…*

typedef struct { int a; float b; my_type c[10];} compound_t

02/21/14 37

Enumeration Datatype Class

( string, value of type Datatype )

Enumeration Datatype

Only Integer datatype for now

Datatype Class

1…*RED 0GREEN 1BLUE 2

1

02/21/14 38

Variable Length Datatype Class

Variable Length Datatype Class

Datatype Class 1

A 1 2 3 4 5

RED GREEN BLUE

C

02/21/14 39

Reading Dataset into Memory from File

File Memory

2D array of integers 3D array of floats

02/21/14 40

Reading Dataset into Memory from File

Memory Dataspace___________________

Selection

Memory Datatype

Dataset Xfer Prp List

Copy of File Dataspace_______________________

Selection Dataset

Dataspace

Datatype

Data

File Memory2D array with selected rectangle

3D array with selected union of cubes

floats

H5DreadBuffer( )

I/O hint