itertree python package

Welcome to itertree python package

Release 1.0.5 - fully released

• Do you have to store data in a tree like structure?
• Do you need good performance and a reach feature set in the tree object?
• You like to serialize and store the structure in files?
• And is it helpful for you if you can link subtrees from other trees and add local items in this "inherited" parts?

Please give itertree package a try!

The main class for construction of the trees is the iTree-class. Here is a simple representation of a itertree:

 iTree('root', value='xyz')
 > iTree('subitem', value='abc')
 > iTree(('tuple', 'tag'), value={'dict': 'value'})
 .  > iTree('subtag', value=1)
 .  > iTree('subtag', value=2)
 > iTree('tag', value=[1, 2, 3])

Every node in the itertree (iTree-object) contains two main parts:

* First the related sub-structure (`iTree`-children)
* Second the item related value attribute were any kind of object can be stored in

The itertree solution can be compared with nested lists or dicts. Other packages that targeting in the in the same direction are anytree, (l)xml.ElementTree, pyTooling.Tree. In detail the feature-set and functional focus of iTree is a bit different.

Installation

Use the package manager pip to install the itertree package.

pip install itertree

The package has no dependencies to other external packages. But the comparison tests with other packages are obviously only possible if the other packages are available.

But we have two recommendations which give the package additional performance:

• blist - Highly recommended! This will speedup the iTree performance in huge trees especially for inserting and lefthandside operations

  • package link: https://pypi.org/project/blist/
  • documentation: http://stutzbachenterprises.com/blist/.

  -> in case the package is not found normal list object will be used instead

  -> If you like to utilize blist under Python 3.10 and 3.11 read the installation hints in the Introduction of the itertree documentation.

• orjson - A quicker json parser that used to create the JSON structures during serializing/deserializing

  -> in case orjson is not found, ujson package is checked too

  -> in case both not found normal json package will be used

Feature Overview

The main features of the itertree package can be summarized with:

• trees can be structured in different levels (nested trees: parent - children - sub-children - ....)
• the identification tag (key) can be any kind of hashable object
• tags must not be unique (same tags are enumerated and collect in a tag-family)
• item access is possible via tag-index-pair, absolute index, slices, index-lists or filters
• the iTree-object keeps the order of the added children
• an iTree-object can contain linked/referenced items (linking to other internal tree parts or to an external itertree file is supported)
• in a linked iTree specific items can be localized and they can cover linked elements (overloading)
• supports standard serialization via export/import to JSON (incl. numpy and OrderedDict data serialization)
• designed for performance (huge trees with hundreds of levels and over a million of items)
• helper functions and data models which can be used to specify the valid values are delivered too
• it's a pure python package (should be easy usable in all environments)
• in general the iTree-class can be seen as a functional mix of lists and dicts with deeper levels and references

Here is very simple example of itertree usage:

>>> from itertree import * # required for all examples shown in the documentation
>>> # Create root item:
>>> root = iTree('root', value={'mykey': 0})
>>> # Append children:
>>> root.append(iTree('sub', value={'mykey': 1}))
iTree('sub', value={'mykey': 1})
>>> root.append(iTree('sub', value={'mykey': 2}))
iTree('sub', value={'mykey': 2})
>>> root.append(iTree('sub', value={'mykey': 3}))
iTree('sub', value={'mykey': 3})
>>> # Show tree content:
>>> root.render()
iTree('root', value={'mykey': 0})
 > iTree('sub', value={'mykey': 1})
 > iTree('sub', value={'mykey': 2})
 > iTree('sub', value={'mykey': 3})

First steps

All important classes of the package are published by the package __init__.py file so that the functionality of itertree can be reached by importing:

>>> from itertree import *

This import is a precondition for all shown code examples in this documentation.

The itertree trees are build by adding iTree-objects to a iTree-parent-object. This means we do not have an external tree generator the tree is build by using the appending functionalities of the objects itself.

We start now building an itertree with the recommended method for adding items append(). The user might use the lazy way via += operator ( __iadd__() ) too. Both operations will add a child item at the end of the parent sub-tree (like append() in lists).

>>> root = iTree('root')  # first we create a root element (parent)
>>> root.append(iTree(tag='child', value=0))  # add a child append method
iTree('child', value=0)
>>> root.append(iTree((1, 2, 3), 1))  # add next child (the given tag is tuple, any hashable object can be used as tag)
iTree((1, 2, 3), value=1)
>>> root += iTree(tag='child2', value=2)  # next child could be added via += operator too
>>> root.render()  # show the created tree
iTree('root')
 > iTree('child', value=0)
 > iTree((1, 2, 3), value=1)
 > iTree('child2', value=2)

Each iTree-object has a tag which is the main part of the identifier of the object. For tags you can use any type of hashable objects.

Different than the keys in dictionaries the given tags must not be unique! The user should understand that in general iTree-objects behave more like nested lists than nested dicts:

>>> root.append(iTree('child', 5))
iTree('child', value=5)
>>> root.append(iTree('child', 6))
iTree('child', value=6)
>>> root.render()
iTree('root')
 > iTree('child', value=0)
 > iTree((1, 2, 3), value=1)
 > iTree('child2', value=2)
 > iTree('child', value=5)
 > iTree('child', value=6)

In the iTree object equal tags are enumerated in a tag-family and they can be targeted via a tag-index-pair (family-tag,family-index). In the "wording" of ìTree this pair is named a key because it is unique like the keys in dicts. To summarize the items in an iTree can be accessed via absolute index (like in lists) or they can be reached by giving the key (tag-index-pair) which is comparable to the key in dicts (both ways are very quick).

>>> print(root['child', 1])  # target via key -> tag_idx pair
iTree('child', value=5)
>>> print(root[3])  # target via absolute index
iTree('child', value=5)

E.g.: To add sub-items we can address the child item also by absolute index and add a sub-item.

>>> root[0].append(iTree('subchild'))
iTree('subchild')
>>> print(root[0][0])
iTree('subchild')

After the tree is generated we can iterate over the tree:

>>> a = [i for i in root]
>>> len(a)
5
>>> print(a)
[iTree('child', value=0, subtree=[iTree('subchild')]), iTree((1, 2, 3), value=1), iTree('child2', value=2), iTree('child', value=5), iTree('child', value=6)]
>>> b = list(root.deep)  # The list is build by iterating over all nested children
>>> len(b)  # The item: root[0][0] is considered in this iteration too
6
>>> print(b)
[iTree('child', value=0, subtree=[iTree('subchild')]), iTree('subchild'), iTree((1, 2, 3), value=1), iTree('child2', value=2), iTree('child', value=5), iTree('child', value=6)]

As shown in the example we have the possibility to iterate over the first level only (children) or we use the internal class absolute index (like in lists) or they can be reached by giving the key (tag-index-pair) which is comparable to the key in dicts (both ways are very quick).

>>> print(root['child', 1])  # target via key -> tag_idx pair
iTree('child', value=5)
>>> print(root[3])  # target via absolute index
iTree('child', value=5)

E.g.: To add sub-items we can address the child item also by absolute index and add a sub-item.

>>> root[0].append(iTree('subchild'))
iTree('subchild')
>>> print(root[0][0])
iTree('subchild')

Many iterable methods have a filter_method parameter in which a filtering method can be placed that targets specific properties of the items.

>>> # ----> filtering method can be placed that targets specific properties of the items.
>>> a = [i for i in root.deep.iter(filter_method=lambda i: type(i.value) is int and i.value % 2 == 0)]  # search even data items
>>> print(a)
[iTree('child', value=0, subtree=[iTree('subchild'), iTree('subchild')]), iTree('child2', value=2), iTree('child', value=6)]

In case no value is given the iTree will take automatically the itertree.NoValue object as value. In case an iTree is instanced without tag the tag value itertree.NoTag will be used.

>>> empty_item = iTree()
>>> print(empty_item)
iTree()
>>> print(empty_item.tag)
<class 'itertree.itree_helpers.NoTag'>
>>> print(empty_item.value)
<class 'itertree.itree_helpers.NoValue'>

At least the itertree can be stored and reconstructed from a file. We can also link an item to a specific item in a file (external link) or create internal links.

>>> root.dump('dt.itz',overwrite=True)  # itz is the recommended file ending for the zipped dataset file
9cd3a9a644af51ea94c82f64ca4ccf745b4a1dd717958beec0cfeb9b0647ba73
>>> root2 = iTree().load('dt.itz')  # loading a iTree from a file
>>> print(root2 == root)
True
>>> root += iTree('link', link=iTLink('dt.itz',[('child', 0)]))  # The node item will integrate the children of the linked item.

iTree-generators vs. lists

As the package name itertree suggests we have several possibilities to iterate over the tree items. The related functions are realized internally via generators. We have generators targeting the children only (level 1) and we have others which ran in-depth into the whole tree structure targeting all the internal items (children, sub-children,...). The provided generators can be easily casted into real iterators via build-in iter()-method (most often the cast is not required, if target method takes generators (uses __iter__())).

If mytree is an iTree-object e.g. you can iterate via:

• iter(mytree) - level 1 iterator over all children delivers the items

• iter(mytree.keys()) - level 1 iterator over all children delivers the tag-idx of the items

• iter(mytree.values()) - level 1 iterator over all children delivers the values of the items

• iter(mytree.items()) - level 1 iterator over all children delivers the (tag_idx,item) pair of the items

• iter(mytree.deep) - flatten iterator over all in-depth items in the tree delivers the items

• iter(mytree.deep.tag_idx_path()) - flatten iterator over all in-depth items delivers the (tag-idx-path,item) pair

• iter(mytree.deep.idx_path()) - flatten iterator over all in-depth items delivers the (abs. index,item) pair

• mytree.get.iter(*target_path) - delivers an iterator over all items targeted via target_path (multi item target)

The usage of generators (iterators) give some big advantages over the usage of lists related to performance and memory consumption. The main idea is to combine all the filtering and iterable options together before you start the final iteration (consume the iter-generator). The instancing of generators/iterators is is very quick and independent from the number of items the object wil iter over. E.g. if the user would casts the inbetween results of multiple operations into 'list'-objects it would take relative long time and the memory consumption would be much more. Therefore it is recommended to build (cascade) all required operations based on the given generator/iterator object. And only at the very end we should consume the generator/iterator. If the code is build like this it is very quick and needs less memory. So please avoid type casts to lists in between the operations. It is very helpful if the user have a look at the powerful itertools-package which can be utilized for those proposes.

If the user really wants to to end-up in a list-object he can easy cast the generator by using the list() statement (The cast might be needed for list related functionalities like len()):

Related to generators/iterators the user should know:

• The StopIteration exception must be handled in case of empty generators.

• An generator can be consumed only one time. To reuse an generator multiple times you may have a look at itertools.tee().

Here are some possible usages of the iteration functions in itertree (imagine large trees for small trees the example operations are equivalent):

>>> myresultlist = list(root.deep)  # this is quick even for huge number of items
>>> first_item = list(root.deep)[0]  # but this is slower (list-type-cast)  as:
>>> first_item = next(iter(root.deep)) # create an iterator from the generator object
>>> fifth_item = list(root.deep)[4]  # and this is slower as:
>>> import itertools
>>> fifth_item = next(itertools.islice(root.deep, 4, None))

iterators vs. lists

We named the package itertree because when ever a iTree operation delivers multiple items the result will be an iterator (and not a list what the user might expect).

Iterators are very powerful objects especially if you have a huge number of items to be iterated over. Iterators can be created very fast and they can be combined. So you can create very effective filter functions. It's recommended to have a look in the powerful itertools and more_itertools packages to combine it with itertree

The main idea is to combine all the filtering and iterator options together before you start the final iteration (consume the iterator), which might at least end up in the expected list. By this mechanism we do at least only one unique iteration over the items and we must not do multiple typecasts and re-iterations in between even when we combine multiple filters.

If the user really wants to create a list he can easy cast the iterator by using the list() statement:

>>> myresultlist=list(root.iter_all()) #  this is quick even for huge number of items
>>> first_item=list(root.iter_all())[0] # Anyway this is much slower than:
>>> first_item=next(root.iter_all())
>>> fifth_item=list(root.iter_all())[4] # and this is much slower than:
>>> import itertools
>>> fifth_item=next(itertools.isslice(root.iter_all(),4,None))

As it is shown in the performance test the operation list() is very quick (less then 0.5 s on 1 million items (depending on you PC)). And using the index access afterwards is a very good readable code. But as shown here there are quicker solutions available on iterators only.

But we see also two downsides related to iterators:

• The StopIteration exception must be handled in case of empty iterators. To make the handling a bit easier iTree delivers in most cases an empty list if we have no match. But in some cases (e.g. filter operations) the user will get an empty iterator and not the empty list. In itree_helpers the user can find a check function for empty iterators that might help in this case: is_iterator_empty(my_iterator).

• The user must also consider that an iterator can be consumed only one time. To reuse an iterator multiple times you may have a look on itertools.tee().

To summarize this chapter:

We decided that the iTree methods should deliver only iterators (and not lists). This is made to give the user the possibility to utilize the whole iterator power afterwards. If he really needs a list (in most cases for index access) he can cast the iterator easy and quick via the list() statement. But if iTree would directly deliver lists by default we would have a performance drop in all itertree filter functions which is not acceptable from our point of view.