Selecting nodes and extracting data

This document details the selection syntax used in the

Selection Filters Syntax

Selection filters are specified in a single string. Each filter is applied on the list of nodes selected by the previous filter. That way the whole tree of nodes can be traversed.

Single selection filter

Selection is done by processing a filter string and generating validator functions. The validators are applied to potential nodes returning True or False. The nodes must be of similar structure to pysjef.Node, in particular, they should contain attributes nodename, parent, attributes, and children to use all functionality of

General structure of filter string is as follows:

nodename[attr1, attr2 <builtin-operator> value, callable_func(attr1, ..., attrn)]
  • specifying nodename selects all nodes for which node.nodename == nodename (see

  • if no nodename is specified, or nodename == "*" than validator based on nodename is not generated

Following in square brackets are attribute filters:

  • attr# is the name of an attribute stored as key in node.attributes dictionary

  • specifying attr1 selects all nodes for which attr1 in node.attributes (see

  • <builin-operator> is one of the builtin operators for selection filters, currently

    only = to check for string equality is implemented

  • specifying attr2 = value selects all nodes that conaint attr2 and attr2 == value (see

  • callable_func is name of a callable function returning bool which expects attributes

    from node.attributes with name attr1, …, attrn

  • last specification selects nodes that have attributes attr1, and …, and attrn than passes values of attr1 etc. to callable_func and checks if it’s True. The callable_func must be passed as a kwarg. (see


nodename and attribute names (attr#) are not case-sensitive.

Any number of attribute filters can be specified within square brackets, but the corresponding validators are not guaranteed to be in the same order. Generally, validators by attribute name come first, followed by validators by builtin operators, followed by validators by callables.

For example, all of the following selections return the same node

>>> from pysjef import Node, select
>>> import math
>>> point = Node()
>>> point.nodename = "coord"
>>> point.attributes.update([('Cartesian', 'true'), ('3D','false'),('x', 0.0)),('y', 0.0)])
>>> select([point], 'coord[cartesian = True, 3D = false]')
>>> select([point], 'coord[x, y]')
>>> select([point], '[3D = false, is_centre(x,y)]', is_centre= lambda x,y: x**2+y**2 < 1.0e-15)

Return by value

Instead of returning a list of nodes, one can return a list of attribute values by ending filter string with .attribute_name

For example, all of the following return x by value

>>> select([point], 'coord.x')
>>> select([point], '[3D = True].x')
>>> select([point], '.x')


Selection filters are separated by slashes ‘/’.


single slash indicates the following filter applies to direct children


double slash indicates the following filter applies to any descendant at any depth

For example, the following takes a collection of graphs and returns all color nodes that are red

>>> select(graphs, 'graph/coord/colour[red=True]')
>>> select(graphs, 'graph//colour[red=True]')

The first example ensure the hierarchy graph <- coord <- color, while the second only restricts the first and last nodes and will search any node in-between until the end of the tree.


Selection string can either start with a filter or a separator


applies the filter on the provided nodelist


applies the filter on the children


applies the filter on the provided nodelist and descendants at any depth

Return Option

by value

using special syntax Return by value


return list of nodes (default)


select the nodes and return a tree view up to Project nodes (see pysjef.tree_view())


select the nodes and return a tree view to root nodes (see pysjef.tree_view())

Full Example