Implementing a Date Range Parser

2016-11-05

Recently, I was tasked with writing an application in C# to scrape a web page on a schedule. Unfortunately, the dates on the web page followed a variety of formats, and there was no easy way to definitively predict which format was going to be used.

To solve this dilemma, I gathered the different formats used in the web page and wrote a component DateRangeParser.cs to abstract away this date format issue.

DateRangeParser accepts the following formats and returns a start date and end date, as shown below:

Sample usage

string dateRangeString = "19 November - 31 December 2016";
ScheduledOutagesDateRangeParser dateRangeParser;

// parsedDates is a list of two dates: start date and end date
List<DateTime> parsedDates = dateRangeParser.ParseDateRangeString(dateRangeString);
return parsedDates;

Implementation

Inspired by machine learning data processing techniques, DateRangeParser begins by creating a word bank and a proximity table.

A word bank is a 1 x n matrix containing each word in the parsed string, where n is the number of words in the string. For example, the string 07 December 2016 has a word bank of size 1 x 3.

A proximity table is an m x n matrix which stores the distance of a word to every other word in the string. Each column i (ranging from 1 to n) corresponds to a word in the string. And each item j in column i (ranging from 1 to m) also corresponds to a word in the string. Therefore, the proximity table for 07 December 2016 is of size 3 x 3 and looks like the following:

Using the word bank and the proximity table, DateRangeParser looks for all dates (i.e. just the 07 in 07 December 2016) in the given string. For each date found, the algorithm looks for the nearest month and nearest year and uses this information to form a complete date.

The component returns a list of two dates — the start date and the end date in that order.

View the entire component here.