Add input file validation before command-line tool execution

[coderabbitai]

Enhancement Request

Currently, command-line tools in mdx2 load input files using loadobj and nxload, which can raise NeXusError when files don't exist or have invalid paths. However, these errors occur after the tool has started execution and created log files, leading to delays before the issue is detected.

Proposed Enhancement

Check the existence and validity of input files before a command-line tool executes its main logic. This would:

• Avoid creating log files when input files are invalid
• Provide immediate feedback to users about file issues
• Not require major changes to existing command-line tools
• Not break the testing framework (which uses fake filenames and mocking)

Suggested Implementation

Implement the validation inside the with_logging decorator using a mechanism to auto-detect whether a parameter is an input file. One approach could be:

• Define a custom type hint class (e.g., InputFile) instead of str for file parameters
• The decorator can introspect the Parameters dataclass to identify fields with this type
• Validate file existence before executing the wrapped function
• Skip validation during testing (potentially via an environment variable or detecting test context)

Example

@dataclass
class Parameters:
    geom: InputFile  # Auto-detected as input file
    data: InputFile  # Auto-detected as input file
    outfile: str     # Regular string, not validated

The with_logging decorator would check that geom and data files exist before proceeding.

Context

• Related PR: Spm/44 prepare new release #45
• Discussion: Spm/44 prepare new release #45 (comment)
• Requested by: @spmeisburger

Current Behavior

m = loadobj('sandbox/map2.nxs','intensity')  # File doesn't exist
# NeXusError: '/Users/steve/dev/mdx2/sandbox/map2.nxs' does not exist

The error is informative but occurs after log file creation and tool initialization.

[coderabbitai]

Path Canonicalization for Duplicate Detection

During review of PR #45, a potential issue was identified in command-line modules that accept multiple input files (e.g., mdx2.scale, and likely others):

Issue: The current duplicate input file detection in Parameters.__post_init__ uses len(set(self.hkl)) to check for duplicates, but this won't catch the same file referenced via different spellings:

• Relative vs absolute paths: data.nxs vs ./data.nxs vs /full/path/data.nxs
• Symlinks pointing to the same file
• Case variations on case-insensitive filesystems

Impact: Duplicate files can lead to incorrect batch assignments and skewed scaling/merging results.

Suggested Solution: When implementing the InputFile type hint (Path objects), apply Path.resolve() early in validation to produce canonical absolute paths. This ensures:

• Robust duplicate detection
• Consistent path handling across all command-line tools
• Proper handling of symlinks and relative paths

Example implementation:

def __post_init__(self):
    # Normalize paths to canonical form for duplicate detection
    self.hkl = [Path(p).resolve() for p in self.hkl]
    
    # Now duplicate detection is robust
    if len(self.hkl) != len(set(self.hkl)):
        raise ValueError("Duplicate input files are not allowed")

Related: This affects any command-line module accepting multiple input files (scale, merge, etc.).

Reference: #45 (comment)