randonneur

Library to apply flexible changes to datasets


License
MIT
Install
pip install randonneur==0.2.1

Documentation

randonneur

https://www.flickr.com/photos/jswg/35681111281/

Keep moving forward.

Randonneur is a library to make changes to life cycle inventory databases. Specifically, randonneur provides the following:

  • A data format for specifying life cycle inventory data transformations
  • Helper functions to create and validate data in this data format
  • Functions to apply the transformations to data

You can use it to re-link your data to the latest version of a background database, to update existing databases with new data, or to perform other data transformations. Randonneur uses JSON files to describe these changes; contrast this with wurst, which can do these manipulations and more, but documents its manipulations in code.

Another important difference with wurst is that randoneur does not have a fixed data schema - the schema is defined in each file.

randonneur does not provide any data itself, but its sister library randonneur_data has data for many common transformations.

Although designed to work with Brightway, this library is not Brightway-specific.

PyPI Status Python Version License: MIT

Python 🐍 CI/CD tests Codecov

pre-commit

Usage

Generic usage pattern

  • Extract a randonneur data migration file, normally from randonneur_data using randonneur_data.Registry().get_file()
  • Extract an inventory database; this can be in the common Brightway inventory format, but you can also roll your own.
  • Apply the data transformation using migrate_edges, optionally specifying the fields used for matching the transformation data, any mappings necessary to make the transformation data schema fit into your data schema, what filters should be applied to the input data (if any), and which verbs (create, replace, update, delete, or disaggregate) you want to apply.
  • Load the modified data back into a suitable data store.

Here's a basic example:

In [1]: import randonneur as rn
   ...: import randonneur_data as rd
   ...:

In [2]: my_lci = [{
   ...:     'name': "my process",
   ...:     'edges': [{
   ...:         'name': 'Xylene {RER}| xylene production | Cut-off, U',
   ...:         'amount': 1.0
   ...:     }]
   ...: }]
   ...:

In [3]: transformed = rn.migrate_edges_with_stored_data(
   ...:     my_lci,
   ...:     'simapro-ecoinvent-3.9.1-cutoff',
   ...:     config=rn.MigrationConfig(fields=['name'])
   ...: )
   ...: transformed
   ...:
Out[3]:
[{'name': 'my process',
  'edges': [{
    'name': 'xylene production',
    'amount': 1.0,
    'filename': '38175dbb-3f48-592c-83f1-c1f667c4b8fd_43c61790-cbeb-493e-8836-279a12ce3e43.spold',
    'location': 'RER',
    'reference product': 'xylene',
    'unit': 'kg'}]}]

In [4]: rn.migrate_edges_with_stored_data(
   ...:     transformed,
   ...:     'ecoinvent-3.9.1-cutoff-ecoinvent-3.10-cutoff',
   ...: )
   ...:
Out[4]:
[{'name': 'my process',
  'edges': [{'name': 'BTX production, from pyrolysis gas, average',
    'amount': 0.11757529360371775,
    'filename': '38175dbb-3f48-592c-83f1-c1f667c4b8fd_43c61790-cbeb-493e-8836-279a12ce3e43.spold',
    'location': 'RER',
    'reference product': 'xylene, mixed',
    'unit': 'kg',
    'allocation': 0.11757529360371775},
   {'name': 'BTX production, from reformate, average',
    'amount': 0.8824247063962822,
    'filename': '38175dbb-3f48-592c-83f1-c1f667c4b8fd_43c61790-cbeb-493e-8836-279a12ce3e43.spold',
    'location': 'RER',
    'reference product': 'xylene, mixed',
    'unit': 'kg',
    'allocation': 0.8824247063962822}]}]

Data format

Migration data is specified in a JSON file as a single dictionary. This file must include the following keys:

We strongly recommend you provide the following optional attributes:

The following properties should follow the data package specification if provided:

  • description
  • sources
  • homepage
  • created

Finally, at least one change type should be included. The change types are:

  • create
  • replace
  • update
  • delete
  • disaggregate

Here are some examples. First, migrating from one ecoinvent biosphere version to another:

{
  "name": "ecoinvent-3.9.1-biosphere-ecoinvent-3.10-biosphere",
  "description": "Data migration file from ecoinvent-3.9.1-biosphere to ecoinvent-3.10-biosphere generated with `ecoinvent_migrate` version 0.2.0",
  "contributors": [
    {
      "title": "ecoinvent association",
      "path": "https://ecoinvent.org/",
      "role": "author"
    },
    {
      "title": "Chris Mutel",
      "path": "https://chris.mutel.org/",
      "role": "wrangler"
    }
  ],
  "created": "2024-07-24T11:38:11.144509+00:00",
  "version": "2.0.0",
  "licenses": [
    {
      "name": "CC-BY-4.0",
      "path": "https://creativecommons.org/licenses/by/4.0/legalcode",
      "title": "Creative Commons Attribution 4.0 International"
    }
  ],
  "graph_context": [
    "edges"
  ],
  "mapping": {
    "source": {
      "expression language": "XPath",
      "labels": {
        "name": "//*:elementaryExchange/*:name/text()",
        "unit": "//*:elementaryExchange/*:unitName/text()",
        "uuid": "//*:elementaryExchange/@elementaryExchangeId"
      }
    },
    "target": {
      "expression language": "XPath",
      "labels": {
        "name": "//*:elementaryExchange/*:name/text()",
        "unit": "//*:elementaryExchange/*:unitName/text()",
        "uuid": "//*:elementaryExchange/@elementaryExchangeId"
      }
    }
  },
  "source_id": "ecoinvent-3.9.1-biosphere",
  "target_id": "ecoinvent-3.10-biosphere",
  "homepage": "https://github.com/brightway-lca/ecoinvent_migrate",
  "replace": [
    {
      "source": {
        "uuid": "90a94ea5-bca4-483d-a591-2e886c0ff47f",
        "name": "TiO2, 54% in ilmenite, 18% in crude ore"
      },
      "target": {
        "uuid": "2f033407-6060-4e1e-868c-9f362d10fdb2",
        "name": "Titanium"
      },
      "conversion_factor": 0.599,
      "comment": "To be modelled as pure elements, the titanium content of titanium dioxide is 0.599."
    }
  ]
}

Name, moving from SimaPro to ecoinvent nomenclature:

{
  "name": "simapro-ecoinvent-3.10-cutoff",
  "description": "Data migration file from SimaPro 9 to ecoinvent-3.10-cutoff generated by PRé and provided via request at https://support.simapro.com/s/contactsupport",
  "contributors": [
    {
      "title": "PRé",
      "path": "https://pre-sustainability.com/",
      "role": "author"
    },
    {
      "title": "Chris Mutel",
      "path": "https://chris.mutel.org/",
      "role": "wrangler"
    }
  ],
  "created": "2024-07-24T10:37:28.350572+00:00",
  "version": "2.0.0",
  "licenses": [
    {
      "name": "CC-BY-4.0",
      "path": "https://creativecommons.org/licenses/by/4.0/legalcode",
      "title": "Creative Commons Attribution 4.0 International"
    }
  ],
  "graph_context": [
    "edges"
  ],
  "mapping": {
    "source": {
      "expression language": "like JSONPath",
      "labels": {
        "identifier": "Process[*].\"Process identifier\".text",
        "name": "Process[*].Products[*].text[0]",
        "platform_id": "Process[*].\"Platform Identifier\""
      }
    },
    "target": {
      "expression language": "XPath",
      "labels": {
        "filename": "concat(//*:activity/@id, '_', //*:intermediateExchange[*:outputGroup = '0' and @amount > 0]/@intermediateExchangeId, '.spold')",
        "name": "//*:activityName/text()",
        "location": "//*:geography/*:shortname/text()",
        "reference product": "//*:intermediateExchange[*:outputGroup = '0' and @amount > 0]/*:name/text()",
        "unit": "//*:intermediateExchange[*:outputGroup = '0' and @amount > 0]/*:unitName/text()"
      }
    }
  },
  "source_id": "SimaPro-9",
  "target_id": "ecoinvent-3.10-cutoff",
  "replace": [
    {
      "source": {
        "identifier": "EI3ARUNI000011519620702",
        "name": "Sawnwood, azobe, dried (u=15%), planed {RER}| market for sawnwood, azobe, dried (u=15%), planed | Cut-off, U",
        "platform_id": "FE069A7D-BB64-4A2E-8B1B-12960BE28887"
      },
      "target": {
        "filename": "151e46e9-70f3-58de-80b3-eb79a90036b0_148b552a-c50b-465e-84f7-367bda16f04a.spold",
        "name": "market for sawnwood, azobe, dried (u=15%), planed",
        "location": "RER",
        "reference product": "sawnwood, azobe, dried (u=15%), planed",
        "unit": "m3"
      }
    }
  ]
}

Finally, normalizing unit abbreviations to full names:

{
    "name": "generic-brightway-units-normalization",
    "description": "Standard units normalization used in most Brightway projects",
    "contributors": [
        {"title": "Chris Mutel", "path": "https://chris.mutel.org/", "role": "author"}
    ],
    "created": "2024-07-25T06:47:10.575370+00:00",
    "version": "1.0.0",
    "licenses": [
        {
            "name": "CC-BY-4.0",
            "path": "https://creativecommons.org/licenses/by/4.0/legalcode",
            "title": "Creative Commons Attribution 4.0 International",
        }
    ],
    "graph_context": ["nodes", "edges"],
    "mapping": {
        "source": {"expression language": "JSONPath", "labels": {"unit": "Node.unit"}},
        "target": {"expression language": "JSONPath", "labels": {"unit": "Node.unit"}},
    },
    "source_id": "bw_interfaces_schemas-1",
    "target_id": "bw_interfaces_schemas-1",
    "homepage": "https://github.com/brightway-lca/bw_interface_schemas",
    "replace": [
        {"source": {"unit": "a"}, "target": {"unit": "year"}},
        {"source": {"unit": "h"}, "target": {"unit": "hour"}},
        {"source": {"unit": "ha"}, "target": {"unit": "hectare"}},
        {"source": {"unit": "hr"}, "target": {"unit": "hour"}},
        {"source": {"unit": "kg"}, "target": {"unit": "kilogram"}},
    ],
}

See the randonneur_data repo for more real-world implementations.

Common database release identifier standard

At Brightcon 2022 we developed the following simple format for common database release identifiers:

<database name>-<version>-<optional modifier>

database name is usually lower case.

Here are some examples:

  • agribalyse-3.1.1
  • forwast-1
  • ecoinvent-3.10-cutoff
  • simapro-9-biosphere

Theory

In normal life cycle assessment practice, we work with a large variety of software and database applications, and often need to harmonize data across these heterogeneous systems. Because many of these systems do not commonly use simple and unique identifiers, we often need to link across systems based on data attibutes. For example, if the name, location, and unit of an input are the same in system A and B, then we can infer that these refer to the same underlying concept.

In the real world it's not so simple. Each player in the LCA data world is trying to give their users a positive experience, but over time this has led to many different terms for the same concept. Some legacy systems restrictions also prevent complete imports, and cause data transformations that are difficult to reverse engineer.

This library defines both a specification for transformation data files which allow different systems to be linked together by harmonizing the matching attributes, and a software-agnostic reference implementation of functions needed to use that format.

Note that not all verbs or graph object types are currently supported by the reference implmentation.

Transformations

Note

Transformations are serialized to JSON. Therefore, only JSON data types are supported.

Configuration

All transformation operations can be configured via a MigrationConfig object. The following can be specified:

mapping: Change the labels in the migrations data to match your data schema. mapping can change the labels in the migration source and target sections. The mapping input should be a dict with keys "source" and "target", and have values of {old_label: new_label} pairs:

migrate_edges(
    graph=[{"edges": [{"name": "foo"}]}],
    migrations={"update": [{"source": {"not-name": "foo"}, "target": {"location": "bar"}}]},
    config=MigrationConfig(mapping={"source": {"not-name": "name"}})
)
>>> [{"edges": [{"name": "foo", "location": "bar"}]}]

node_filter: A callable which determines whether or not the given node should be modified. Applies to both verbs and edges, with the exception of node creation - it doesn't make sense to filter existing nodes as we are creating new objects.

node_filter needs to be a callable which takes a node object and returns a boolean which tells if the node should be modified. In this example, the filter returns False and the node isn't modified:

migrate_edges(
    graph=[{"edges": [{"name": "foo"}]}],
    migrations={"update": [{"source": {"name": "foo"}, "target": {"location": "bar"}}]},
    config=MigrationConfig(node_filter=lambda node: node.get("sport") == "🏄‍♀️")
)
>>> [{"edges": [{"name": "foo"}]}]

edge_filter: A callable which determines whether or not the given edge should be modified. Applies only to edge transformations, and does not apply to edge creation, as this function is always called on the edge to modified, not on the transformation object. Returns

edge_filter needs to be a callable which takes an edge object and returns a boolean which indicates if the edge should be modified.

fields: A list of object keys as strings, used when checking if the given transformation matches the node or edge under consideration. In other words, only use the fields in fields when checking the source values in each transformation for a match. Each field in fields doesn't have to be in each transformation.

migrate_edges(
    graph=[{"edges": [{"name": "foo"}]}],
    migrations={"update": [
        {"source": {"name": "foo", "missing": "🔍"}, "target": {"location": "bar"}}
    ]},
    config=MigrationConfig(fields=["name"]),
)
>>> [{"edges": [{"name": "foo", "location": "bar"}]}]

verbose: Display progress bars and more logging messages.

edges_label: The label used for edges in the nodes of the graph. Defaults to "edges". In other data formats, this could be "flows" or "exchanges".

migrate_edges(
    graph=[{"e": [{"name": "foo"}]}],
    migrations={"update": [{"source": {"name": "foo"}, "target": {"location": "bar"}}]},
    config=MigrationConfig(edges_label="e"),
)
>>> [{"edges": [{"name": "foo", "location": "bar"}]}]

verbs: The list of transformation types from migrations to apply. Transformations are run in the order as given in verbs, and in some complicated cases you may want to keep the same verbs but change their order to get the desired output state. In general, such complicated transformations should be broken down to smaller discrete and independent transformations whenever possible, and logs checked carefully after their application.

The default value of verbs are the "safe" transformations - replace, update, and disaggregate. To get create and delete you need to specify them in the configuration.

Only the verbs create, disaggregate, replace, update, and delete are used in our functions, regardless of what is given in verbs, as we don't know how to handle custom verbs. We need to write custom functions for each verb as they have difference behaviour.

case_sensitive: Flag indicating whether to do case sensitive matching of transformations to nodes or edges in the graph. Default is false, as practical experience has shown us that cases get commonly changed by software developers or users. Only applies to string values.

migrate_edges(
    graph=[{"edges": [{"name": "foo"}]}],
    migrations={"update": [{"source": {"name": "FOO"}, "target": {"location": "bar"}}]},
    config=MigrationConfig(case_sensitive=False),
)
>>> [{"edges": [{"name": "foo", "location": "bar"}]}]

add_extra_attributes: Flag indicating whether to include additional attributes when doing replace, update, and disaggregate changes. Extra attributes are defined outside the "source" and "target" transformation keys. Note that keys in randonneur.utils.EXCLUDED_ATTRS are never added.

migrate_edges(
    graph=[{"edges": [{"name": "foo"}]}],
    migrations={"update": [{
        "source": {"name": "FOO"},
        "target": {"location": "bar"},
        "comment": "Reason for change",
    }]},
    config=MigrationConfig(add_extra_attributes=True),
)
>>> [{"edges": [{"name": "foo", "location": "bar", "comment": "Reason for change"}]}]

Replace and Update

Warning

Be careful with nested data. The reference implementation takes a relatively simplistic approach, and completely overwrites existing data if a new target value is provided. If you had {'foo': {'nested': {'has_permission': true, 'author': 'someone'}}}, and the transformation gave a new value for nested, it would completely replace the nested dictionary instead of modifying the existing and adding new elements.

replace indicates that a given object should be replaced with a new object. The replacement could substitutes an object one-to-one; as such, the new exchange must be completely defined. Please bear in mind that we are providing transformations for the object that the edge is referring to, not the edge itself. Therefore, the amount, uncertainty, etc. of the edge should not be specified. It the edge amount needs to be rescaled, for example because of a unit conversion, specify a conversion_factor in addition to the source and target.

If allocation is not given, a default value of 1.0 is used.

Aside from the quantitative values, no other data from the original exchange is taken over to the new exchange. If you only want to change a few fields, use an update instead. If you don't want the exchange amount re-scaled, use a combination of delete and create.

The data format for replace type is:

{
    "replace": [{
        "source": {
            # All fields needed to identify the exchange to be replaced
        },
        "target": {
            # All fields needed to define the new exchange
        }
    }]
}

update changes attributes the same way that replace does - the only difference is that replace shows the intent to refer to a new object instead of an existing object with different attributes. Given the messiness of real-world data (i.e. what is truly a new object versus the same object with different descriptions), there is no real bright line between these concepts, and their code implementation is identical.

Create

Creates a new edge or node.

Note

Because this application pattern is so different compared to updating existing values, we don't normally recommend using this functionality. It's preferable to go through the normal data importation process instead.

Note

The reference implementation does not check if creation would create multiple identical objects, and the randonneur specification does not define what should be done if such creation were indicated.

Because we are specifying a new node or exchange, we need to list all information needed to define that object, including the edges and edge amount values. This is different than the other modification types, where relative amounts are given with the key conversion_factor or allocation. We can't give relative amounts here because we have no edge to refer to, and we don't have a surefire way to identify the reference production edge (and there might not be one in any case).

If you want to add an edge to all datasets, or a node to the graph:

{
    "create": [{
        "target": {
            # All fields needed to define the object
        }
    }]
}

To add multiple nodes or edges, add multiple {'target': {}} dictionaries to the create list.

Delete

Delete exchanges. Follows the same patterns as replace and update:

{
    "delete": [{
        "source": {
            # All fields needed to identify the exchange to be deleted
        }
    }]
}

Disaggregate

Disaggregation is splitting one exchange into many. The allocation field is used to determine how much of the exchange passes to each new exchange.

Note

allocation fields do not have to sum to one.

The new exchanges start as copies of the original exchange, and are updating using the additional data provided.

The data format includes a list of new exchanges for each matched source:

{
    "disaggregate": [{
        "source": {
            # All fields needed to identify the exchange to be disaggregated
        },
        "targets": [{
            # Some fields which you want to change
        }]
    }]
}

Contributing

Contributions are very welcome. To learn more, see the Contributor Guide.

License

Distributed under the terms of the MIT license, randonneur is free and open source software.

Issues

If you encounter any problems, please file an issue along with a detailed description.