def_file

Read, write, and modify *.def file from the IPSL/modipsl project.

The *.def file contains model parameters in the key-value format. The format is extremely simple in comparison to similar formats, like *.ini or *.toml, as it doesn't provide sections, nor standarized datatypes. Usually, model configuration files contain dozens or hundred of parameters with scalars (int, float, str), arrays, or special _AUTO_/_AUTOBLOCK_ values.

The special "auto" values are used by the IPSL/modipsl/libIGCM projects to mark place where an external script, called component driver, should inject configuration values.

Examples usage:

// Follows Python convention of the JSON module with load(), dump(), etc. functions
from ipsl_common.modipsl.def_file import load, dump
with open("run_dynamico.def", "r") as f:
    parameters = load(f)
    // Modify loaded parameters
    parameters["start_file_name"] = "start2024.nc"
    with open("new_run_dynamico.def", "w") as g:
        dump(parameters, g)

Loading an example *.def file:

INCLUDEDEF=run_lmdz.def
INCLUDEDEF=run_dynamico.def
use_forcing=y
g=_AUTO_: DEFAULT=9.8
start_file_name=start2023
physics="always"

Gives the following dictionary:

{
    'INCLUDEDEF': ['run_lmdz.def', 'run_dynamico.def'],
    'g': ('_AUTO_', 9.8),
    'physics': '"always"',
    'start_file_name': 'start2023',
    'use_forcing': True
}

Loaded configuration can be altered and subsequently dumped onto a file or into a string. The configuration is easy to view and modify, because it is directly decoded into a Python dictionary. The decoding and encoding process is managed internally by DefFileDecoder and DefFileEncoder classes with decode() and encode() methods.

Classes

DefFileDecoder

DefFileDecoder(include_positions: bool = False)

Bases: Transformer

Decoder performs translation from *.def file to a dictionary.

The translation rules are:

`*.def`	Python	Comment
key-value	`dict`	Including many key-value pairs and `INCLUDEDEF`
array	`list`	Collection of at least 2 elements
string	`str`	Unquoted and quoted (single or double) strings
integer number	`int`	Standard integer values
real number	`float`	Including scientific notation
`true`/`y`	`True`	Case insensitive
`false`/`n`	`False`	Case insensitive
`_AUTO_`	`tuple`	With optional default value
`_AUTOBLOCKER_`	`tuple`	With optional default value

The first step of the decoder is parsing of *.def file. For this task Lark Earley parser is used with a simple grammar expressed with EBNF notation. The *.def grammer doesn't parse well with LALR(1) parser. The parsing produces a parse tree.

The second step transforms the parse tree into Python dictionary using translation rules mentioned in the above table. This transformation is based on a automated visitor pattern called Transformer, which produces the dictionary in a bottom-up manner.

Examples:

from ipsl_common.modipsl.def_file import DefFileDecoder
dictionary = DefFileDecoder().decode(text)

If text contains this *.def file:

radius=6.371229E6
g=9.80665
omega=_AUTO_: DEFAULT=7.292E-5

Then, Python dictionary would look as follows:

{
    "radius": 6.371229e6,
    "g": 9.80665,
    "omega": ("_AUTO_", 7.292e-05),
}

Tip

By default, the result dictionary contains no information about textual layout of the file. However, by using the argument include_positions=True, it is possible to refine the dictionary with exact start/end positions of each value as follows:

{
    "radius": {"value": 6371229.0, "start_pos": 50, "end_pos": 60},
    "g": {"value": 9.80665, "start_pos": 99, "end_pos": 106},
    "omega": {"value": ("_AUTO_", 7.292e-05), "start_pos": 158, "end_pos": 182},
}

Initialize DefFileDecoder.

Parameters:

include_positions
(bool, default: False ) –

include textual positions of values (offset from the start)

Source code in ipsl_common/modipsl/def_file.py

def __init__(self, include_positions: bool = False) -> None:
    """Initialize DefFileDecoder.

    Args:
        include_positions: include textual positions of values (offset from the start)
    """
    self._include_positions = include_positions

Methods:

decode

decode(content: str) -> dict

Decode *.def content into dictionary.

Parameters:

content
(str) –

content of the *.def file

Returns:

dict ( dict ) –

Decoded *.def file

Source code in ipsl_common/modipsl/def_file.py

def decode(self, content: str) -> dict:
    """Decode `*.def` content into dictionary.

    Args:
        content (str): content of the `*.def` file

    Returns:
        dict: Decoded `*.def` file
    """
    parse_tree = self._parser.parse(content)
    return self.transform(parse_tree)

DefFileEncoder

DefFileEncoder(
    truthy_value: str = "true", falsey_value: str = "false"
)

Encoder performs translation from a dictionary to *.def file.

The translation rules are:

Python	`*.def`	Comment
`dict`	key-value	Each `INCLUDEDEF` value translates to a single key-value
`list`	array	Of at least 2 elements
`str`	string	Quoted strings will contain explicit quote characters
`int`	integer number	---
`float`	real number	Including scientific notation
`bool`	`true`/`false`	Can be specified with encode arguments
`tuple`	`_AUTO_`/`_AUTOBLOCKER_`	With optional default value at second position in tuple

The translation is straightforward, based on Python type a specific conversion is performed. No grammar, nor parse tree is used during this step.

Example:

from ipsl_common.modipsl.def_file import DefFileEncoder
text = DefFileEncoder().encode(dictionary)

If dictionary contains:

{
    "radius": 6.371229e6,
    "g": 9.80665,
    "omega": ("_AUTO_", 7.292e-05),
}

Then, the encoded *.def file would look as follows:

radius = 6371229.0
g = 9.80665
omega = _AUTO_: DEFAULT=7.292e-05

Tip

Python representation of the *.def file doesn't contain any textual position of particular elements (keys, values, comments, whitespaces, etc.), thus, re-encoding of the exact input *.def file is impossible. In order to recreate the original file, or modify a file while keeping the original comments, whitespaces, and order of elements, use the designated modify functions.

Initialize DefFileEncoder.

Parameters:

truthy_value
(str, default: 'true' ) –

label used to encode True
falsey_value
(str, default: 'false' ) –

label used to encode False

Source code in ipsl_common/modipsl/def_file.py

def __init__(
    self,
    truthy_value: str = "true",
    falsey_value: str = "false",
):
    """Initialize DefFileEncoder.

    Args:
        truthy_value: label used to encode True
        falsey_value: label used to encode False
    """
    self._truthy_value = truthy_value
    self._falsey_value = falsey_value

Methods:

encode

encode(obj: object) -> str

Encode dictionary or other Python object into *.def file.

This function works not only on full decoded *.def files, but it also work on particular Python object such as a list, or a tuple. In such case, it will take the given object and apply one of the encoding rules mentioned before.

Parameters:

obj
(object) –

dictionary or Python object

Returns:

str –

Encoded text of a *.def file

Source code in ipsl_common/modipsl/def_file.py

def encode(self, obj: object) -> str:
    """Encode dictionary or other Python object into `*.def` file.

    This function works not only on full decoded `*.def` files,
    but it also work on particular Python object such as a list,
    or a tuple. In such case, it will take the given object and apply
    one of the encoding rules mentioned before.

    Args:
        obj: dictionary or Python object

    Returns:
        Encoded text of a `*.def` file
    """
    # bool must be tested before int, because it is a subclass of int class.
    if isinstance(obj, bool):
        return self._truthy_value if obj else self._falsey_value
    elif isinstance(obj, int | float):
        return str(obj)
    elif isinstance(obj, tuple):
        if len(obj) != 2:
            raise ValueError(
                f"Only tuples with len=2 are encoded. Invalid tuple: {obj}"
            )
        auto, default = obj
        if auto not in _AUTO_VALUES:
            raise ValueError(
                f"First element of a tuple must be one of {_AUTO_VALUES}. Invalid tuple: {obj}"
            )
        if not self._is_scalar(default):
            raise TypeError(
                f"Second element of the tuple must be a scalar: bool, int, float, str, or None. Invalid tuple: {obj}"
            )
        if default is not None:
            return f"{auto}: DEFAULT={self.encode(default)}"
        else:
            return str(auto)
    elif isinstance(obj, dict) and len(obj) == 1:
        # Using tuple assignment
        ((k, v),) = obj.items()
        if k == _INCLUDEDEF:
            return self._encode_include(v)
        else:
            return self._encode_entry(k, v)
    elif isinstance(obj, dict):
        return "\n".join([self.encode({k: v}) for k, v in obj.items()])
    elif isinstance(obj, list):
        if len(obj) < 2:
            raise ValueError("List must have at least two elements")
        if not all(map(self._is_scalar, obj)):
            raise TypeError(
                f"All array values must be simple scalars: bool, int, float, str, or None. Instead got: {obj}"
            )
        return ", ".join(map(self.encode, obj))
    else:
        return str(obj)

Functions:

dump

dump(obj: dict, buffer: TextIOBase) -> None

Dump dictionary into *.def text/file buffer.

Parameters:

obj
(dict) –

dictionary to dump to a file
buffer
(TextIOBase) –

text or file buffer for storing *.def file

Source code in ipsl_common/modipsl/def_file.py

def dump(obj: dict, buffer: TextIOBase) -> None:
    """Dump dictionary into *.def text/file buffer.

    Args:
        obj: dictionary to dump to a file
        buffer: text or file buffer for storing *.def file
    """
    if not buffer.writable():
        raise ValueError("Text buffer (TextIOBase) must be writable")
    buffer.write(dumps(obj))

dumps

dumps(obj: dict) -> str

Dump dictionary into *.def string.

Parameters:

obj
(dict) –

dictionary to dump to a file

Source code in ipsl_common/modipsl/def_file.py

def dumps(obj: dict) -> str:
    """Dump dictionary into *.def string.

    Args:
        obj: dictionary to dump to a file
    """
    return DefFileEncoder().encode(obj)

load

load(
    buffer: TextIOBase,
    include_positions: bool = False,
    return_text: bool = False,
) -> dict | tuple[dict, str]

Load *.def text/file buffer into dictionary.

Parameters:

buffer
(TextIOBase) –

text or file buffer with the *.def file
include_positions
(bool, default: False ) –

include textual positions of values

Returns:

dict ( dict | tuple[dict, str] ) –

Loaded *.def file

Source code in ipsl_common/modipsl/def_file.py

def load(
    buffer: TextIOBase, include_positions: bool = False, return_text: bool = False
) -> dict | tuple[dict, str]:
    """Load `*.def` text/file buffer into dictionary.

    Args:
        buffer: text or file buffer with the `*.def` file
        include_positions: include textual positions of values

    Returns:
        dict: Loaded `*.def` file
    """
    if not buffer.readable():
        raise ValueError("Text buffer (TextIOBase) must be readable")
    text = buffer.read()
    if return_text:
        return loads(text, include_positions), text
    else:
        return loads(text, include_positions)

loads

loads(text: str, include_positions: bool = False) -> dict

Load *.def file string into dictionary.

Parameters:

text
(str) –

content of the *.def file
include_positions
(bool, default: False ) –

include textual positions of values

Returns:

dict ( dict ) –

Loaded *.def file

Source code in ipsl_common/modipsl/def_file.py

def loads(text: str, include_positions: bool = False) -> dict:
    """Load `*.def` file string into dictionary.

    Args:
        text: content of the `*.def` file
        include_positions: include textual positions of values

    Returns:
        dict: Loaded `*.def` file
    """
    return DefFileDecoder(include_positions).decode(text)

modify

modify(
    buffer: TextIOBase,
    new_obj: dict,
    buffer_out: TextIOBase | None = None,
    insert_header: str = "",
) -> None

Modify *.def text/file buffer with minimal amount of changes.

THe output text/file buffer follows changes made to new_obj representation of *.def file. The modifications are performed, so that the minimal amount of changes is applied. As a result, the diff between the old and new file content is minimal and no comments or whitespaces are lost beyond what is neccessary. The new_obj can remove, modify, and/or add new key-value pairs.

Parameters:

buffer
(TextIOBase) –

text or file buffer for reading the file (optionally to write to the file)
new_obj
(dict) –

modified representation of the *.def file content
buffer_out
(TextIOBase | None, default: None ) –

optional output buffer for writing the modified file content
insert_header
(str, default: '' ) –

optional header inserted before appended key-value pairs

Source code in ipsl_common/modipsl/def_file.py

def modify(
    buffer: TextIOBase,
    new_obj: dict,
    buffer_out: TextIOBase | None = None,
    insert_header: str = "",
) -> None:
    """Modify `*.def` text/file buffer with minimal amount of changes.

    THe output text/file buffer follows changes made to `new_obj` representation of `*.def` file.
    The modifications are performed, so that the minimal amount of changes is applied.
    As a result, the diff between the old and new file content is minimal and no comments
    or whitespaces are lost beyond what is neccessary. The `new_obj` can remove, modify,
    and/or add new key-value pairs.

    Args:
        buffer(TextIOBase): text or file buffer for reading the file (optionally to write to the file)
        new_obj(dict): modified representation of the *.def file content
        buffer_out(TextIOBase | None): optional output buffer for writing the modified file content
        insert_header(str): optional header inserted before appended key-value pairs
    """
    # Buffer must always be readable
    if not buffer.readable():
        raise ValueError("Text buffer (TextIOBase) must be readable")
    # Buffer must be writeable if buffer_out is not passed
    if buffer_out is None and not buffer.writable():
        raise ValueError("Text buffer (TextIOBase) must be writeable")
    # Otherwise, buffer_out must be writeable
    elif buffer_out is not None and buffer_out.writable():
        raise ValueError("Text buffer_out (TextIOBase) must be writeable")

    target_buffer = buffer_out if buffer_out is not None else buffer
    target_buffer.write(modifys(buffer.read(), new_obj, insert_header=insert_header))

modifys

modifys(
    text: str, new_obj: dict, insert_header: str = ""
) -> str

Modify *.def file string with minimal amount of changes.

THe output text follows changes made to new_obj representation of *.def file. The modifications are performed, so that the minimal amount of changes is applied. As a result, the diff between the old and new file content is minimal and no comments or whitespaces are lost beyond what is neccessary. The new_obj can remove, modify, and/or add new key-value pairs.

Parameters:

text
(str) –

input text with *.def content to modify
new_obj
(dict) –

modified representation of the *.def file content
insert_header
(str, default: '' ) –

optional header inserted before appended key-value pairs

Returns:

str ( str ) –

Modified text

Source code in ipsl_common/modipsl/def_file.py

def modifys(text: str, new_obj: dict, insert_header: str = "") -> str:
    """Modify `*.def` file string with minimal amount of changes.

    THe output text follows changes made to `new_obj` representation of `*.def` file.
    The modifications are performed, so that the minimal amount of changes is applied.
    As a result, the diff between the old and new file content is minimal and no comments
    or whitespaces are lost beyond what is neccessary. The `new_obj` can remove, modify,
    and/or add new key-value pairs.

    Args:
        text(str): input text with *.def content to modify
        new_obj(dict): modified representation of the *.def file content
        insert_header(str): optional header inserted before appended key-value pairs

    Returns:
        str: Modified text
    """
    old_obj = loads(text, include_positions=True)
    encoder = DefFileEncoder()
    replacements = []

    def _eat_newline(text: str, position: int) -> int:
        """Return new position with the newline eaten"""
        return position + 1 if text[position] == "\n" else position

    # Replace removed key-value pairs by empty strings
    deleted_keys = old_obj.keys() - new_obj.keys()
    for k in deleted_keys:
        v = old_obj[k]
        replacements.append((v["key_start_pos"], _eat_newline(text, v["end_pos"]), ""))

    # TODO: refactor this part in the future, now I don't have enough strength to do so
    include_insert_pos = 0
    old_includes_with_pos = old_obj.pop("INCLUDEDEF", [])
    old_includes = [m["value"] for m in old_includes_with_pos]
    to_remove = old_includes.copy()
    for m in new_obj.pop("INCLUDEDEF", []):
        if m in old_includes:
            to_remove.remove(m)
            include_insert_pos = (
                old_includes_with_pos[old_includes.index(m)]["end_pos"] + 1
            )
        else:
            replacements.append(
                (include_insert_pos, include_insert_pos, f"INCLUDEDEF={m}\n")
            )
    # Whatever has left is to remove
    for m in to_remove:
        d = old_includes_with_pos[old_includes.index(m)]
        replacements.append((d["key_start_pos"], _eat_newline(text, d["end_pos"]), ""))

    # Apply modification to existing keys
    for k, v in old_obj.items():
        if k in new_obj and v["value"] != new_obj[k]:
            replacements.append(
                (v["start_pos"], v["end_pos"], encoder.encode(new_obj[k]))
            )
    new_text = replace_text(text, replacements)
    new_text += insert_header
    # Then, add new keys at the end
    # INFO: We cannot use .keys() to find the new keys, because this method
    # returns a set which doesn't keep keys order. Instead, we can iterate
    # over items in dictionaries which is guaranteed to preserve the
    # insertion order since Python 3.7
    # (https://docs.python.org/3/library/stdtypes.html#typesmapping)
    new_keys = []
    for k in new_obj:
        if k not in old_obj:
            new_keys.append(k)

    if append_keys := encoder.encode({k: new_obj[k] for k in new_keys}):
        new_text += append_keys
        new_text += "\n"
    return new_text

def_file

Classes

DefFileDecoder

`include_positions`

Methods:

decode

`content`

DefFileEncoder

`truthy_value`

`falsey_value`

Methods:

encode

`obj`

Functions:

dump

`obj`

`buffer`

dumps

`obj`

load

`buffer`

`include_positions`

loads

`text`

`include_positions`

modify

`buffer`

`new_obj`

`buffer_out`

`insert_header`

modifys

`text`

`new_obj`

`insert_header`