在C＃中解析日期/时间的最宽松方法？

Question

I'm parsing a broad range of RSS feeds - apprently they all use their own way to show the timestamp of the article.

Now we even found one that uses a local words, like Donderdag 17 juli 2018 .

At the moment we have a fallback mechanism where we just fall back to DateTime.UtcNow when we can't parse the date.

Still I would like to make a best attempt. What is the best way to really loosely parse a DateTime in C#? So it can handle ie:

• 13-11-2018 14.32
• donderdag 13 november 2018, 14:32
• 13 nov 2018
• 14:32 13.11.2018
• 2018-11-13T16:32:00+2:00

etc. I know that this would not be foolproof, but still I like to make a best attempt.

Is there any recommended way? Or do I have to roll my own?

Answer 1

You could use DateTime.TryParseExact and include all the expected formats .

DateTime result;
if( DateTime.TryParseExact(input, new [] {"dd-MM-yyyy HH.mm", "dddd dd MMMM yyyy, HH:mm", "more formats here"}, CultureInfo.CreateSpecificCulture("nl-NL"), DateTimeStyles.None, out result)) {
  Console.WriteLine("Succeeded " + result);
}

---

The only big "gotcha" here is date formats where the date and month are in ambiguous positions. I do not see any in your example but if you were to mix cultures in one stream then it could become a problem. As an example the US generally starts a formatted date with the month while the Netherlands starts it with the day of the month. If this is a problem there is no way to handle this dynamically in your use case above unless you also get the culture from the RSS stream in which case you could try to create a set of culture specific parsing rules.

Answer 2

This suggestion is not specific to date times, but you could try to use parser combinators , especially if you decide to roll your own solution. There are multiple libs for .net, Sprache for example.

Answer 3

Loosely parsing date times from mixed sources if data is probably not a good idea. Some things like Microsoft's text-to-speech may try, but it can sometimes have the effect of reading consecutive dates as

October first, November first, December first, January thirteenth, etc.

The only way loose parsing can be made somewhat reliable is if one can use other cues to associate dates with whatever wrote them. If you have a bunch of dates that occur at the top level of a particular feed, and you find that all parsing patterns that work for all of them yield the same results, then it's likely that that parsing pattern is parsing the dates correctly. The biggest parts of such an endeavor, however, will likely not be parsing the dates, but rather (1) ensuring that dates that are written in different formats get grouped separately, and (2) providing a means by which an operator can assist the program in places where it has trouble.

Incidentally, I don't know if any date parsing programs make use of attached weekdays as part of format validation, but they could often help. For example, "2-1-2018" could either be January 2 or February 1, but "Thursday 2-1-2018" could only be the latter. It may be helpful when parsing numeric dates from a source whose format isn't fully established to determine what the weekday would be with each method of parsing and check whether the input contains something that looks like a weekday matching one but not the other.

Answer 4

You can use the TryParse method to try to parse the strings, while looping through all cultures to capture any culture differences in the string. The following method will parse all standard formats for all cultures and return the date in the out parameter if it's found.

Note that the danger here is that some dates will have ambiguous month and day values (any number less than 13 could be a month or a day). In that case, the result will be the first culture found that matches, which may not be correct.

Here's the code:

public static bool TryParseAllCultures(string formattedDate, 
    out DateTime result)
{
    // First try in our local culture
    if (DateTime.TryParse(formattedDate, out result)) return true;

    foreach (var cultureInfo in CultureInfo.GetCultures(CultureTypes.AllCultures))
    {
        if (DateTime.TryParse(formattedDate, cultureInfo, DateTimeStyles.None, 
            out result))
        {
            return true;
        }
    }

    return false;
}

---

Sample usage

Note: I modified one of your dates because the date itself was invalid! The second date used to be "donderdag 13 november 2018", except the 13th is dienstag (Tuesday), not donderdag (Thursday).

private static void Main()
{
    DateTime date;

    var dateFormats = new List<string>
    {
        "13-11-2018 14.32",
        "donderdag 15 november 2018, 14:32",
        "13 nov 2018",
        "14:32 13.11.2018",
        "2018-11-13T16:32:00+2:00"
    };

    DateTime result;

    foreach (var dateFormat in dateFormats)
    {
        if (TryParseAllCultures(dateFormat, out result))
        {
            Console.ForegroundColor = ConsoleColor.Green;
            Console.WriteLine($"SUCCESS: {dateFormat.PadRight(36, '.')} {result}");
        }
        else
        {
            Console.ForegroundColor = ConsoleColor.Red;
            Console.WriteLine($"ERROR: Unable to parse format: {dateFormat}");
        }

        Console.ResetColor();
    }

    GetKeyFromUser("\nDone! Press any key to exit...");
}

Output