Convert human readable string to TimeSpan

Question

I want to pass time interval as a string (to program args) and then I want to parse it to TimeSpan. I created simple class:

public static class TimeSpanConverter
 {
 public static TimeSpan Convert(string input)
 {
 var units = new Dictionary<string, int>()
 {
 {@"(\d+)(ms|mili[|s]|milisecon[|s])", 1 },
 {@"(\d+)(s|sec|second[|s])", 1000 },
 {@"(\d+)(m|min[|s])", 60000 },
 {@"(\d+)(h|hour[|s])", 3600000 },
 {@"(\d+)(d|day[|s])", 86400000 },
 {@"(\d+)(w|week[|s])", 604800000 },
 };
 var timespan = new TimeSpan();
 foreach(var x in units)
 {
 var matches = Regex.Matches(input, x.Key);
 foreach(Match match in matches)
 {
 var amount = System.Convert.ToInt32(match.Groups[1].Value);
 timespan = timespan.Add(TimeSpan.FromMilliseconds(x.Value * amount));
 }
 }
 return timespan;
 }
 }

Which is working, and I can easly parse like this:

var interval = TimeSpanConverter.Convert("1day13hour2min12s52ms");

What do you think about that? How can I improve that? I have a feeling, that it is not well coded.

Answer 1

public static class TimeSpanConverter
 {
 public static TimeSpan Convert(string input)

To people who use WPF a lot, the name hints at this class being a TypeConverter or an IValueConverter. I would say that it's really a parser. Similarly, I think that the method should be called Parse.

---

 var units = new Dictionary<string, int>()
 {
 {@"(\d+)(ms|mili[|s]|milisecon[|s])", 1 },
 {@"(\d+)(s|sec|second[|s])", 1000 },
 {@"(\d+)(m|min[|s])", 60000 },
 {@"(\d+)(h|hour[|s])", 3600000 },
 {@"(\d+)(d|day[|s])", 86400000 },
 {@"(\d+)(w|week[|s])", 604800000 },
 };

The prefix milli has two ls. millisecond ends in a d.

[|s] is more idiomatically written as s?.

Given the way the regexes are used (searching for matches without anchoring), most of them could be simplified. E.g. any match for min would also match m, so @"(\d+)m" will find exactly the same multiples of a minute.

The numbers on the right are magic numbers, and at a glance I can't be completely certain that they're correct. I would prefer to use TimeSpan instances and take advantage of TimeSpan.FromMilliseconds(1), TimeSpan.FromSeconds(1), etc.

I think that a real world example of a human-readable string would be "1 day, 13 hours, 2 minutes, 12 seconds, and 52 ms". At the very least, I would add \s* between the digits and the units.

Another quite realistic example would be "2 minutes and 13.5 seconds".

It's also plausible that people will use : as a separator: "2 days, 11:04:20"

Then you get ugly stuff like months and years, which don't have fixed lengths.

If you want to refuse to parse some of these, that's fine. But in that case, I think the method still needs to handle them by throwing an exception. At present they would all return TimeSpan.Zero, which is misleading.

Have you given any thought to localisation? If you only want to use this in an English-language context, wouldn't [0-9] make more sense than \d? If you're matching all digits, you should have test cases for things like "1day١hour1minute1second".

---

 var timespan = new TimeSpan();

Can you think of a more descriptive name for this variable?

---

 var matches = Regex.Matches(input, x.Key);
 foreach(Match match in matches)

If the string has more than one match for a given regex, isn't that rather worrying? I wouldn't consider "4hours2minutes1hour" to be a very realistic input.

---

 var amount = System.Convert.ToInt32(match.Groups[1].Value);
 timespan = timespan.Add(TimeSpan.FromMilliseconds(x.Value * amount));

Again, amount isn't a very descriptive name.

TimeSpan supports the operator +, and IMO total += addend; is more readable than total = total.Add(addend);.

• t3chb0t: I would add \s* between the digits and the units. - In such cases I usually normalize a string first by replacing all multiple-whitespaces by a single one to avoid using the *.

Answer 2

Adding to Peter Taylors answer:

@"(\d+)(ms|mili[|s]|milisecon[|s])"

@"(\d+)(m|min[|s])"

"5ms" will be caught by both patterns above.

---

var amount = System.Convert.ToInt32(match.Groups[1].Value);

For weeks > 3 this will make an overflow when multiplied with 604800000.

You should use double.Parse(), because TimeSpan.FromMilliseconds() takes a double as argument.

Answer 3

I don't have experience in programming in C#, but I do have some regex advice to share.

You can DRY your pattern by

(\d+)(?:ms|mili(?:secon)?s?)
(\d+)(?:s(?:ec)?|seconds?)
(\d+)(?:m|mins?)
(\d+)(?:h|hours?)
(\d+)(?:d|days?)
(\d+)(?:w|weeks?)

I am encouraging:

• non-capturing groups when there is no use for the matched substrings.
• I am using the zero or more quantifier (?) to reduce repetition and alternation (pipes).
• the omission of pipes in your character classes because they are useless in your expected input.
• s? to make pluralization optional.

If you need to ensure that m is not followed by s when executing the minutes pattern, you will need to extend the pattern. Perhap m(?!s)|mins?.

If you are in control of the incoming strings, please correct the spelling of milliseconds