Fix sporadic parsing errors with binary ScalarList

[Copilot]

Binary-format scalarList files would sporadically raise FoamFileDecodeError, UnicodeDecodeError, or ValueError depending on the specific float values written — because the parser tried to interpret raw binary bytes as ASCII text before falling back to the binary parser.

Root causes

• _parse_token fatal errors escaping suppression: When binary data begins with an ASCII letter followed by ( (e.g. H(), _parse_token's depth-tracking logic scanned through the binary payload and raised FoamFileDecodeError (not a ParseError subclass), bypassing all contextlib.suppress(ParseError) guards. Same issue for UnicodeDecodeError when a binary byte matching " triggered the quoted-string path.

• _ASCIINumericListParser decode/reshape errors: The comment-aware regex (_pattern, no re.ASCII flag) could match binary data containing non-ASCII bytes, causing UnicodeDecodeError from data.decode("ascii"). For vector lists, a mismatched element count caused ValueError from reshape().

• int32 binary parser masking float64: Binary parsers were chained — int32 was tried first, float64 only on failure. The byte 0x29 ()) can appear at offset count × 4 within float64 data, making the int32 parser falsely succeed and float64 never be attempted.

Changes

• _parse_token: Raise ParseError instead of FoamFileDecodeError for unmatched parens and unclosed quoted/#{} tokens; wrap all decode() calls to convert UnicodeDecodeError → ParseError.

• _ASCIINumericListParser.__call__ and _parse_ascii_faces_like_list: Wrap data.decode("ascii") and ret.reshape() to convert UnicodeDecodeError/ValueError → ParseError.

• _parse_binary_numeric_list: Wrap ret.reshape() to convert ValueError → ParseError.

• _parse_standalone_data_entry: Try all three binary dtypes (int32, float64, float64×3) independently rather than in a chain, always keeping the result with the highest consumed position. Float64 will always win over int32 for the same count because it consumes twice the bytes. Also catch FoamFileDecodeError/UnicodeDecodeError from _parse_data.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Parsing error with binary ScalarList</issue_title>
<issue_description>I experienced a strange parsing error when trying to read a scalarList with the FoamFile class when in binary format, in the form:

[...]
  File "/home/framognino/anaconda3/envs/TabICE/lib/python3.11/site-packages/foamlib/_files/_parsing/_parser.py", line 1143, in parse_located
    raise e.make_fatal() from None
foamlib._files._parsing.exceptions.foamlib.FoamFileDecodeError: parsing failed on line 83, column 264:
���d�?��$I]��?�4��J�?��s���?����)�?�|l�7��?�}r���?�������?Y�K�j,�?4������?�� M���?������?��1D�?to�m�?@b��g��?BP�t�?i!���e�?���>�?���,�?����B�?a�o���?�8�����?�%����?��e��?�r�����?�Vsw���?�x����?5�Pa�   �?k�T_s0�?��N�%W�?�uA����?Vmn����?,�b���?)`K�U6�?J<'��4�?+�X-�?V���-�?�$�mZ��?���,���?��>�vz�?3��NA�?E��        �?�J�+\��|4��?
                                                                                                                                                      c�?���?�,�?��72���?$�&����?�l�?�R��9�?nSm�'�?�?C�?�bB_���?d~�9���?��wdX~�?ak�&�H�?^bN�>�?s�}:���?�Y@7f��?W�|7%s�?���>�?��T�3       �?�2a����?`�s�`��?"<(
                                                                                                                                                                                                                                                                       ^
Expected: keyword or standalone data

This error is really sporadic. I can tell that it is not caused by floating point special cases (I had datasets with nan and inf and worked fine). I tried storing the data in the FoamFile as list and numpy arrays before saving but got still the same problem.

Thank you a lot in advance and let me know if you need further info.

How to reproduce

Here are the functions that I used to write the scalar list, and the one that I am using to reading it.

import os
from typing import Iterable
from foamlib import FoamFile
import numpy as np
import traceback
import struct

# Here I define some convenience functions to I/O scalar fields with the foamlib FoamFile class
def readScalarList(fileName:str) -> Iterable[float]:
    """
    Reads an OpenFOAM file storing a scalarList. Automatically detects if the file is binary or not.

    Args:
        fileName (str): Name of the OpenFOAM file.
    
    Raises:
        IOError: If the file does not exist or if it does not store a scalarList
    
    Returns:
        Iterable[float]: The data stored in the file.
    """
    #Check path:
    if not(os.path.isfile(fileName)):
        raise IOError("File '{}' not found.".format(fileName))
    
    with FoamFile(fileName) as f:
        if f.class_ != "scalarList":
            raise IOError("File '{}' does not store a scalarList.".format(fileName))
        
        data = f[None]
        
        # Check if the data were correctly read as float, otherwise cast them
        if isinstance(data, np.ndarray) and data.dtype == np.int64:
            data.dtype = np.float64
        
        return data

#Write OF file with scalar list
def writeScalarList(values:Iterable[float], path:str, *, overwrite:bool=False, binary:bool=False) -> None:
    """
    Write an OpenFOAM file storing a scalarList.

    Args:
        values (Iterable[float]): The data to store.
        path (str): The location where to file the scalarList.
        overwrite (bool, optional): Overwrite if found? Defaults to False.
        binary (bool, optional): Write in binary? Defaults to False.
    
    Raises:
        IOError: If the file exists and overwrite is False.
    """
    #Check path:
    if os.path.isfile(path) and not overwrite:
        raise IOError("File '{}' exists. Run with overwrite=True.".format(path))
    
    if os.path.exists(path):
        os.remove(path)

    #Create the file object
    with FoamFile(path) as File:
        File.path.touch()
        root, file = os.path.split(path)
        # Use add bacause this prevents trying to parse the file which would lead to an error since it is empty
        File.add("FoamFile", {
            "class":"scalarList",
            "version":2.0,
            "object":file,
            "location":os.path.split(os.path.abspath(root))[1],
            "format": "binary" if binary else "ascii"
        })
        File.format = "binary" if binary else "ascii"
        File.add(None,np.array(values).flatten().tolist())

#---------------------------------------------------
# Testing with foamlib
print("Testing with foamlib")

# Loading ascii file 
data = readScalarList("data_ascii")

# Writing in binary format
writeScalarList(data, "data_binary", overwrite=True, binary=True)

# Read again in binary
try:
    data_binary = readScalarList("data_binary")
    print("Done!")
except BaseException as e:
    print("Error!")
    traceback.print_exc()

Output:

Testing with foamlib
Error!
Traceback (most recent ...

</details>



<!-- START COPILOT CODING AGENT SUFFIX -->

- Fixes gerlero/foamlib#760

<!-- START COPILOT CODING AGENT TIPS -->
---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our [2 minute survey](https://gh.io/copilot-coding-agent-survey).