read functions should return insightful errors

[tlukkezen]

read functions should return insightful errors

[RDWimmers]

I see two options now, both breaking changes:

• Require more content about the file argument (e.g. with a string_content=["path","file"] argument, but that would only have a meaning for file arguments of type str (kinda ugly)
• As pandas.read_csv() does: Accept str | io.StringIO types for file and always infer the str type as a Path, and the io.StringIO as file content. This would require the user to convert a string to StringIO object

I like the second one. we can set the type of filepath_or_buffer to Path | io.StringIO. then we dont use strings and can use the Path.is_file() method.

[tlukkezen]

In the current implementation, we just try to parse the possible contents of file and if it fails we move on to the next. The last parsing option that fails (parsing as bro-xml by default) is the one that provides the error. This is random behaviour and is the reason that the error that is returned doesn't reflect the actual problem.

I see the following resolutions now:

• Require more content about the file argument (e.g. with a string_content=["path","file"] argument, but that would only have a meaning for file arguments of type str (kinda ugly) (Breaking change)

• As pandas.read_csv() does: Accept str | io.StringIO types for file and always infer the str type as a Path, and the io.StringIO as file content. This would require the user to convert a string to StringIO object. (Breaking change)

• Return a general custom exception (e.g. CPTParsingError) when all parsing options have failed. Although this will still return a ValueError for an erroneous gef-file path.

[tversteeg]

When a string has the form of a path separated by slashes or even a single short word it can never be a valid XML or GEF file right? All XML files need to start with a < character, so that's easy, and all GEF files have the form of key: value, so I think a proper heuristic would be:

Is almost certainly path if:

1. Is the length <= 255 characters
2. Does it end with case-insensitive .gef or .xml
3. Does it not contain any < or : characters
4. Can it be opened from disk