# HDF5 in Python: PyTables

HDF5 is a wonderful file format you can use to put into tons of data with easy, without the need to think about endianess, binary formats and so on.
Pytables is an extremely optimized library built on top of HDF5 capabilities to make even simpler the use of this type of file.
It’s also possible to navigate into a file graphically with ViTables.
Here I would like to present some of the features I use more often.

Open, flush and close a file

import tables as tb

h5 = tb.openfile("filename.h5", 'r')
...
h5.flush()
h5.close()


where 'r' means “open the file in read-only mode”. It’s also possible to open it as 'w' (create a new file: it overwrites the file if it still exists) and 'a' (append: create if it does not exist, if it exists, read and modify it).

Create a group to contain some data

group = h5.createGroup(h5.root, "group")


where h5.root is the location of the new object created (where we want to create the group) and can also be passe as string (“/”) and "group" is the string with the name.

Store an array

a = np.array([....])
array = h5.createArray(h5.root.group, 'name', array_to_store, 'title')
array = h5.createArray("/group", 'name', array_to_store, 'title')

f = open("filename.csv", 'r')
values = np.genfromtxt(StringIO(line), dtype=([('column_1', 'i8'), ('column_2', 'f4'), ('column_3', 'f4')]), delimiter=',')
values.shape = 1

# or in an equivalent way, if the file dimensions permit to lad the entire file:
# values = np.genfromtxt("filename.csv", dtype=([('column_1', 'i8'), ('column_2', 'f4'), ('column_3', # 'f4')]), delimiter=',')

h5 = tb.openFile('filename.h5', 'w')
table = h5.createTable(h5.root, description=values, name='table_name', title="table_description", expectedrows=12158536)
table.flush()

for line in f:
values = np.genfromtxt(StringIO(line), dtype=([('column_1', 'i8'), ('column_2', 'f4'), ('column_3', 'f4')]), delimiter=',')
values.shape = 1
table.append(values)

table.flush()
h5.flush()
h5.close()
f.close()


It’s also possible to walk all the nodes under a group:

for i in h5.walkNodes(h5.root.group):
print i


and to delete a node/array/table:

h5.removeNode(h5.root.group, 'name')


All the nodes are available through their path:

h5.getNode("/group", "name")
h5.root.group.name


to read a table or an array, you can use the function read()

h5.getNode("/group", "name").read()