formulas implements an interpreter for Excel formulas, which parses and compile Excel formulas expressions.

Moreover, it compiles Excel workbooks to python and executes without using the Excel COM server. Hence, Excel is not needed.

To install it use (with root privileges):

$ pip install formulas

Or download the last git version and use (with root privileges):

$ python setup.py install

Some additional functionality is enabled installing the following extras:

• excel: enables to compile Excel workbooks to python and execute using: :class:`~formulas.excel.ExcelModel`.
• plot: enables to plot the formula ast and the Excel model.

To install formulas and all extras, do:

$ pip install formulas[all]

To help with the testing and the development of formulas, you can install the development version:

$ pip install https://github.com/vinci1it2000/formulas/archive/dev.zip

The formulas command-line interface works with spreadsheet models and accepts .xlsx, .ods, and .json inputs.

A typical workflow starts by calculating a workbook. You can override input values directly from the command line and request specific cells to be rendered in the output.

$ formulas calc test/test_files/excel.xlsx \
    --overwrite "'[excel.xlsx]'!INPUT_A=3" \
    --overwrite "'[excel.xlsx]DATA'!B3=1" \
    --render "'[excel.xlsx]DATA'!C2=result" \
    --output-format json

Spreadsheet models can also be converted into a portable JSON representation. This is useful when the model needs to be versioned, inspected, or executed without the original workbook.

$ formulas build test/test_files/excel.xlsx \
    --output-file model.json

For validation purposes, a workbook can be tested directly from the CLI. The following command runs the tests and prints a short summary.

$ formulas test test/test_files/excel.xlsx --summary

Finally, a model can be exposed as a lightweight HTTP API, allowing other applications to execute it remotely. The serve command requires the optional web dependencies (pip install formulas[web]).

$ formulas serve test/test_files/excel.xlsx \
    --host 127.0.0.1 \
    --port 5000

Each command provides additional options and examples through the built-in help system:

$ formulas COMMAND --help

The following sections will show how to:

• parse a Excel formulas;
• load, compile, and execute a Excel workbook;
• extract a sub-model from a Excel workbook;
• add a custom function.

An example how to parse and execute an Excel formula is the following:

>>> import formulas
>>> func = formulas.Parser().ast('=(1 + 1) + B3 / A2')[1].compile()

To visualize formula model and get the input order you can do the following:

.. dispatcher:: func
   :opt: graph_attr={'ratio': '1'}
   :code:

    >>> list(func.inputs)
    ['A2', 'B3']
    >>> func.plot(view=False)  # Set view=True to plot in the default browser.
    SiteMap({=((1 + 1) + (B3 / A2)): SiteMap({})})

Finally to execute the formula and plot the workflow:

.. dispatcher:: func
   :opt: workflow=True, graph_attr={'ratio': '1'}
   :code:

    >>> func(1, 5)
    Array(7.0, dtype=object)
    >>> func.plot(workflow=True, view=False)  # Set view=True to plot in the default browser.
    SiteMap({=((1 + 1) + (B3 / A2)): SiteMap({})})

An example how to load, calculate, and write an Excel workbook is the following:

.. testsetup::

    >>> import os.path as osp
    >>> from setup import mydir
    >>> fpath = osp.join(mydir, 'test/test_files/excel.xlsx')
    >>> dir_output = osp.join(mydir, 'test/test_files/tmp')

>>> import formulas
>>> fpath, dir_output = 'excel.xlsx', 'output'  # doctest: +SKIP
>>> xl_model = formulas.ExcelModel().loads(fpath).finish()
>>> xl_model.calculate()
Solution(...)
>>> xl_model.write(dirpath=dir_output)
{'EXCEL.XLSX': {Book: <openpyxl.workbook.workbook.Workbook ...>}}

To plot the dependency graph that depict relationships between Excel cells:

.. dispatcher:: dsp
   :code:

    >>> dsp = xl_model.dsp
    >>> dsp.plot(view=False)  # Set view=True to plot in the default browser.
    SiteMap({ExcelModel: SiteMap(...)})

To overwrite the default inputs that are defined by the excel file or to impose some value to a specific cell:

>>> xl_model.calculate(
...     inputs={
...         "'[excel.xlsx]'!INPUT_A": 3,  # To overwrite the default value.
...         "'[excel.xlsx]DATA'!B3": 1  # To impose a value to B3 cell.
...     },
...     outputs=[
...        "'[excel.xlsx]DATA'!C2", "'[excel.xlsx]DATA'!C4"
...     ] # To define the outputs that you want to calculate.
... )
 Solution({"'[excel.xlsx]'!INPUT_A": <Ranges>('[excel.xlsx]DATA'!A2)=[[3]],
           "'[excel.xlsx]DATA'!B3": <Ranges>('[excel.xlsx]DATA'!B3)=[[1]],
           "'[excel.xlsx]DATA'!A2": <Ranges>('[excel.xlsx]DATA'!A2)=[[3]],
           "'[excel.xlsx]DATA'!A3": <Ranges>('[excel.xlsx]DATA'!A3)=[[6]],
           "'[excel.xlsx]DATA'!A4": <Ranges>('[excel.xlsx]DATA'!A4)=[[5]],
           "'[excel.xlsx]DATA'!D2": <Ranges>('[excel.xlsx]DATA'!D2)=[[1]],
           "'[excel.xlsx]'!INPUT_B": <Ranges>('[excel.xlsx]DATA'!A3)=[[6]],
           "'[excel.xlsx]'!INPUT_C": <Ranges>('[excel.xlsx]DATA'!A4)=[[5]],
           "'[excel.xlsx]DATA'!A3:A4": <Ranges>('[excel.xlsx]DATA'!A3:A4)=[[6] [5]],
           "'[excel.xlsx]DATA'!B2": <Ranges>('[excel.xlsx]DATA'!B2)=[[9.0]],
           "'[excel.xlsx]DATA'!D3": <Ranges>('[excel.xlsx]DATA'!D3)=[[2.0]],
           "'[excel.xlsx]DATA'!C2": <Ranges>('[excel.xlsx]DATA'!C2)=[[10.0]],
           "'[excel.xlsx]DATA'!D4": <Ranges>('[excel.xlsx]DATA'!D4)=[[3.0]],
           "'[excel.xlsx]DATA'!C4": <Ranges>('[excel.xlsx]DATA'!C4)=[[4.0]]})

To build a single function out of an excel model with fixed inputs and outputs, you can use the compile method of the ExcelModel that returns a DispatchPipe. This is a function where the inputs and outputs are defined by the data node ids (i.e., cell references).

.. dispatcher:: func
   :code:

    >>> func = xl_model.compile(
    ...     inputs=[
    ...         "'[excel.xlsx]'!INPUT_A",  # First argument of the function.
    ...         "'[excel.xlsx]DATA'!B3"   # Second argument of the function.
    ...     ], # To define function inputs.
    ...     outputs=[
    ...         "'[excel.xlsx]DATA'!C2", "'[excel.xlsx]DATA'!C4"
    ...     ] # To define function outputs.
    ... )
    >>> func
    <schedula.utils.dsp.DispatchPipe object at ...>
    >>> [v.value[0, 0] for v in func(3, 1)]  # To retrieve the data.
    [10.0, 4.0]
    >>> func.plot(view=False)  # Set view=True to plot in the default browser.
    SiteMap({ExcelModel: SiteMap(...)})

Alternatively, to load a partial excel model from the output cells, you can use the from_ranges method of the ExcelModel:

.. dispatcher:: dsp
   :code:

    >>> xl = formulas.ExcelModel().from_ranges(
    ...     "'[%s]DATA'!C2:D2" % fpath,  # Output range.
    ...     "'[%s]DATA'!B4" % fpath,  # Output cell.
    ... )
    >>> dsp = xl.dsp
    >>> sorted(dsp.data_nodes)
    ["'[excel.xlsx]'!INPUT_A",
     "'[excel.xlsx]'!INPUT_B",
     "'[excel.xlsx]'!INPUT_C",
     "'[excel.xlsx]DATA'!A2",
     "'[excel.xlsx]DATA'!A3",
     "'[excel.xlsx]DATA'!A3:A4",
     "'[excel.xlsx]DATA'!A4",
     "'[excel.xlsx]DATA'!B2",
     "'[excel.xlsx]DATA'!B3",
     "'[excel.xlsx]DATA'!B4",
     "'[excel.xlsx]DATA'!C2",
     "'[excel.xlsx]DATA'!D2"]

The ExcelModel can be exported/imported to/from a readable JSON format. The reason of this functionality is to have format that can be easily maintained (e.g. using version control programs like git). Follows an example on how to export/import to/from JSON an ExcelModel:

.. testsetup::

    >>> import formulas
    >>> import os.path as osp
    >>> from setup import mydir
    >>> fpath = osp.join(mydir, 'test/test_files/excel.xlsx')
    >>> xl_model = formulas.ExcelModel().loads(fpath).finish()

>>> import json
>>> xl_dict = xl_model.to_dict()  # To JSON-able dict.
>>> xl_dict  # Exported format. # doctest: +SKIP
{
 "'[excel.xlsx]DATA'!A1": "inputs",
 "'[excel.xlsx]DATA'!B1": "Intermediate",
 "'[excel.xlsx]DATA'!C1": "outputs",
 "'[excel.xlsx]DATA'!D1": "defaults",
 "'[excel.xlsx]DATA'!A2": 2,
 "'[excel.xlsx]DATA'!D2": 1,
 "'[excel.xlsx]DATA'!A3": 6,
 "'[excel.xlsx]DATA'!A4": 5,
 "'[excel.xlsx]DATA'!B2": "=('[excel.xlsx]DATA'!A2 + '[excel.xlsx]DATA'!A3)",
 "'[excel.xlsx]DATA'!C2": "=(('[excel.xlsx]DATA'!B2 / '[excel.xlsx]DATA'!B3) + '[excel.xlsx]DATA'!D2)",
 "'[excel.xlsx]DATA'!B3": "=('[excel.xlsx]DATA'!B2 - '[excel.xlsx]DATA'!A3)",
 "'[excel.xlsx]DATA'!C3": "=(('[excel.xlsx]DATA'!C2 * '[excel.xlsx]DATA'!A2) + '[excel.xlsx]DATA'!D3)",
 "'[excel.xlsx]DATA'!D3": "=(1 + '[excel.xlsx]DATA'!D2)",
 "'[excel.xlsx]DATA'!B4": "=MAX('[excel.xlsx]DATA'!A3:A4, '[excel.xlsx]DATA'!B2)",
 "'[excel.xlsx]DATA'!C4": "=(('[excel.xlsx]DATA'!B3 ^ '[excel.xlsx]DATA'!C2) + '[excel.xlsx]DATA'!D4)",
 "'[excel.xlsx]DATA'!D4": "=(1 + '[excel.xlsx]DATA'!D3)"
}
>>> xl_json = json.dumps(xl_dict, indent=True)  # To JSON.
>>> xl_model = formulas.ExcelModel().from_dict(json.loads(xl_json))  # From JSON.

An example how to add a custom function to the formula parser is the following:

>>> import formulas
>>> FUNCTIONS = formulas.get_functions()
>>> FUNCTIONS['MYFUNC'] = lambda x, y: 1 + y + x
>>> func = formulas.Parser().ast('=MYFUNC(1, 2)')[1].compile()
>>> func()
4

Formulas can also be embedded as a calculation engine inside lightweight applications and automated workflows, without requiring Excel or another spreadsheet GUI.

This example loads a workbook once, exposes it through a Flask application, and calls the JSON API through a test client.

from formulas.app import create_app

app = create_app(files=('test/test_files/excel.xlsx',), circular=False)
client = app.test_client()
response = client.post('/api/calculate', json={
    'inputs': {
        "'[excel.xlsx]'!INPUT_A": 3,
        "'[excel.xlsx]DATA'!B3": 1,
    },
    'renders': ["'[excel.xlsx]DATA'!C2=result"],
})

assert response.status_code == 200
assert response.get_json()['outputs'] == {'result': 10.0}

This example creates a temporary batch file and runs formulas calc over two scenarios.

import json
import subprocess
import sys
import tempfile
from pathlib import Path

with tempfile.TemporaryDirectory() as tmp:
    batch = Path(tmp) / 'batch.json'
    batch.write_text(json.dumps([
        {
            'name': 'base',
            'overwrite': {
                "'[excel.xlsx]'!INPUT_A": 3,
                "'[excel.xlsx]DATA'!B3": 1,
            },
            'renders': ["'[excel.xlsx]DATA'!C2=result"],
        },
        {
            'name': 'stress',
            'overwrite': {
                "'[excel.xlsx]'!INPUT_A": 4,
                "'[excel.xlsx]DATA'!B3": 1,
            },
            'renders': ["'[excel.xlsx]DATA'!C2=result"],
        },
    ], indent=2))

    result = subprocess.run([
        sys.executable, '-m', 'formulas.cli', 'calc',
        'test/test_files/excel.xlsx',
        '--batch', str(batch),
        '--processes', '2',
        '--output-format', 'json',
        '--output-dir', tmp,
    ], capture_output=True, text=True, check=False)

    assert result.returncode == 0, result.stderr
    summary = json.loads(result.stdout)
    assert [item['name'] for item in summary] == ['base', 'stress']

This example treats a workbook as a transformation step over structured input records.

import formulas

model = formulas.ExcelModel().loads('test/test_files/excel.xlsx').finish()
func = model.compile(
    inputs=["'[excel.xlsx]'!INPUT_A", "'[excel.xlsx]DATA'!B3"],
    outputs=["'[excel.xlsx]DATA'!C2"],
)
records = [
    {'id': 'row-1', 'input_a': 3, 'b3': 1},
    {'id': 'row-2', 'input_a': 4, 'b3': 1},
]
results = []

for record in records:
    result, = func(record['input_a'], record['b3'])
    results.append({
        'id': record['id'],
        'result': result.value[0, 0],
    })

assert results == [
    {'id': 'row-1', 'result': 10.0},
    {'id': 'row-2', 'result': 11.0},
]

The current Excel function coverage is tracked in the test workbook test/test_files/test.xlsx, sheet COVERAGE. The table below summarizes the current implementation status by category.

Overall coverage is currently 483 out of 536 functions (90.1%).

Things yet to do: implement the missing Excel formulas.