Yeah. When the alternative is manual.

The code I posted and that I could extrapolate out a 90% success rate was not including a rules engine that could possibly push this rate to 95% or more.

Anytime that you have data that is not strictly formatted, the heuristics get complicated and there will be errors. There is no way to avoid that but you can attempt to mitigate it as best you can as the alternative, manual reconciliation, is worse.

I'm trying to help Tracy out by giving her a decent starting position. You will note that she needs to identify every component of the address such as prefix, suffix, et al. This is not easy.

If you think 90% is "pathetic", you are welcome to your opinion. But until you post a better alternative, in code, STFU.




> 90% success rate on addresses is pathetic. Is there any case where 90% success rate on parsing would be acceptable?