Broadly, it's actually easy to see how naive parsing can become quadratic if a programmer isn't careful. You parse by processing a long string/file via "get-a-

Broadly, it's actually easy to see how naive parsing can become quadratic if a programmer isn't careful. You parse by processing a long string/file via "get-a-token(), add-token-to-structure()" repeating until you have built your structure or made an error in the construction process (here a token a kind, a type of substring and get-a-token retrieves the substring type in some form). The problems come because the naive programmer is only thinking about their structure building process, so they use some library function to get their tokens. Whether that's a regular expression interpreter or some other flexible facility by provided by the library, it's easy for it's definition of a token-string to include things "not in the string" as well as things in the string - a "quoted string" begins and ends with a quote, so to discover that X isn't a quoted string, you have to scan to the end. So if thing go wrong, each call to get-a-token may scan the while file to make sure this token isn't that. Just an example. The problem here seems to be tokenizing floats, which also has a "where it end" problem in c++. And the alternative is parsing floats yourself, which indeed no one wants to do since they are complex and defined by a complex standard.