Merging two identical trees with different attributes

[chhei]

Example Code

# Pseudo code to illustrate what I'd like to do

from bigtree import dict_to_tree
path_dict_1 = {
    "a": {"age": 90},
    "a/b": {"age": 65},
    "a/c": {"age": 60},
    "a/b/d": {"age": 40},
    "a/c/f": {"age": 38},
    "a/b/e/h": {"age": 6},
    }

path_dict_2 = {
    "a": {"colour": "green"},
    "a/b": {"colour": "pink"},
    "a/c": {"colour": "ultraviolet"},
    "a/b/d": {"colour": "cadmiumorange"},
    "a/c/f": {"colour": "froggygreen"},
    "a/b/e/h": {"colour": "wombatbrown"},
    }
root_1 = dict_to_tree(path_dict_1)
root_2 = dict_to_tree(path_dict_2)

copy_nodes_from_tree_to_tree(
    from_tree=root_1,
    to_tree=root_2,
    from_paths=["a"],
    to_paths=["a"],
    merge_children=True,
    merge_attributes=True # <--- something like this would be desireable
)

root_2.show(attribute_list=['age', 'colour'])

>>> a [age=90, colour=green]
>>> ├── b [age=65, colour=pink]
>>> │   ├── d [age=40, colour=cadmiumorange]
>>> │   └── e
>>> │       └── h [age=6, colour=wombatbrown]
>>> └── c [age=6, colour=ultraviolet]
>>>     └── f [age=38, colour=froggygreen]

Description

Hi - first off thanks for this excellent module!

I have a question to which I was not able to find a simple answer for: is it possible to directly merge two identical tree structures (same nodes/hierarchies) which have different attributes in a way that the merged tree simply contains a merged/union of the attributes of both trees?

I guess the other option would be to iterate over the two trees and construct a new merged tree or do this without bigtree in some sort of pre-processing step.

Thanks!

(edited initial typos)

Additional Context

No response

[kayjan]

Hello, thanks for the suggestion! In this scenario it will be more straightforward to convert the trees to dataframe, perform dataframe operation of merging, then converting it back to a tree because there is no 'shifting' or 'copying' involved if it is just to combine the attributes.

Something like this -

import pandas as pd
from bigtree import dict_to_tree, tree_to_dataframe, dataframe_to_tree

# Your provided input
path_dict_1 = {
    "a": {"age": 90},
    "a/b": {"age": 65},
    "a/c": {"age": 60},
    "a/b/d": {"age": 40},
    "a/c/f": {"age": 38},
    "a/b/e/h": {"age": 6},
    }

path_dict_2 = {
    "a": {"colour": "green"},
    "a/b": {"colour": "pink"},
    "a/c": {"colour": "ultraviolet"},
    "a/b/d": {"colour": "cadmiumorange"},
    "a/c/f": {"colour": "froggygreen"},
    "a/b/e/h": {"colour": "wombatbrown"},
    }

root_1 = dict_to_tree(path_dict_1)
root_2 = dict_to_tree(path_dict_2)

# Suggested processing
df1 = tree_to_dataframe(root_1, all_attrs=True)
df2 = tree_to_dataframe(root_2, all_attrs=True)
df_merge = pd.merge(df1, df2, on=["path", "name"])
root_merge = dataframe_to_tree(df_merge)
root_merge.show(all_attrs=True)

This will result in the tree structure

a [age=90.0, colour=green]
├── b [age=65.0, colour=pink]
│   ├── d [age=40.0, colour=cadmiumorange]
│   └── e
│       └── h [age=6.0, colour=wombatbrown]
└── c [age=60.0, colour=ultraviolet]
    └── f [age=38.0, colour=froggygreen]

Also, in your provided example I fixed some typos where it should be path_dict_1 and not path_dict1 and also I assume in path_dict_2 you meant the attribute colour instead of age. Let me know if this approach solves your use case!

[chhei]

Thanks @kayjan (also for fixing those typos) - the intermediate step of using pandas to merge works beautifully and (at least for now) does what I want to do.

[kayjan]

No problem! Will keep this feature in mind for the shift/copy functions because currently there are merge_children/merge_leaves/overriding hyperparameters(?) as well, will see how merge_attribute fit into the various modes. 😄

[kayjan]

Just an update to this, have implemented merge_attribute parameter in v0.23.0. Do perform pip install --upgrade bigtree. For your scenario, this will be how it is used:

from bigtree import copy_nodes_from_tree_to_tree, dict_to_tree

# Your provided input
path_dict_1 = {
    "a": {"age": 90},
    "a/b": {"age": 65},
    "a/c": {"age": 60},
    "a/b/d": {"age": 40},
    "a/c/f": {"age": 38},
    "a/b/e": {"age": 35},  # adding this so age column can be all int
    "a/b/e/h": {"age": 6},
    }

path_dict_2 = {
    "a": {"colour": "green"},
    "a/b": {"colour": "pink"},
    "a/c": {"colour": "ultraviolet"},
    "a/b/d": {"colour": "cadmiumorange"},
    "a/c/f": {"colour": "froggygreen"},
    "a/b/e/h": {"colour": "wombatbrown"},
    }

root_1 = dict_to_tree(path_dict_1)
root_2 = dict_to_tree(path_dict_2)

# Processing
copy_nodes_from_tree_to_tree(
    from_tree=root_1,
    to_tree=root_2,
    from_paths=["a"],
    to_paths=["a"],
    merge_attribute=True,
    merge_children=True,
)

root_2.show(all_attrs=True)

This will result in the tree structure

a [age=90, colour=green]
├── b [age=65, colour=pink]
│   ├── d [age=40, colour=cadmiumorange]
│   └── e [age=35]
│       └── h [age=6, colour=wombatbrown]
└── c [age=60, colour=ultraviolet]
    └── f [age=38, colour=froggygreen]

[chhei]

@kayjan - this is awesome. Many thanks for implementing this, really appreciated! I've tested this quickly with one of my scenarios and it works like a charm. I'll keep testing and will report back in case I run into issues.