Incorrect parsing of USPC for 1-digit classes

[Radcliffe]

Issue

US patent classifications having single-digit classes are parsed incorrectly. For example, 2/322 is misinterpreted as 23/22.

Explanation

In the source XML files, the US patent classification is represented by a character string of variable length. The first three characters contain the class, and the remaining characters contain the subclass. If the class has less than three characters then it is padded with leading spaces. In particular, if the class has only one digit, then the string should have two leading spaces.

The string containing the US patent classification is cleaned using a number of functions, including xml_util.remove_escape_sequences(string). This function replaces the two leading spaces with a single space, causing the string to be parsed incorrectly.