libgen.py: IndexError('basic_string::substr: __pos (which is 4294967295) > this->size() (which is 45)')

[tmplt]

Querying bookwyrm -t 'mass effect' yields an pybind11::error_already_set during self.bookwyrm.feed(item). While the core::item is constructed, the std::string returned from core::get_string(py::dict &, const char*) apparently triggers this; stepping through the code the value is successfully indexed from the dict and the py::str and std::string constructed. The exception is thrown during function return (?).

Originally posted by @tmplt in #75 (comment)

[tmplt]

To be more specific, the item we feed is constructed from the first entry in these search results. The exception is thrown when calling get_string(dict, "title"), which is upon its first occurance in the nonexact_ts constructor.

[tmplt]

Constructing the title field by default yields no exceptions from subsequent calls to get_string(). Does the string contain something illegal?

[tmplt]

This exception does not occur during item construction. It is thrown somewhere in item::matches(const item&, const unsigned int) const! Going by the error message, somewhere we do a bad substr with pos = 2^32 - 1 which logically yields an std::out_of_range. This exception is then translated to a Python IndexError that is then caught in the module runner function.

[tmplt]

The exception is thrown in fuzzywuzzy's fuzz::partial_ratio, lib/fuzzywuzzy/src/fuzzywuzzy.cpp:46 (the only occurance of std::string::substr() where in

size_t long_start = utils::max(0u, block.dpos - block.spos);
size_t long_end   = long_start + shorter.length();
auto long_substr = longer.substr(long_start, long_end);

long_start has been set to 4294967295.