Date copy-constructs given an existing Date, I think type-checking vs throwing away a temporary is a toss-up performance wise. Until you start providing custom formats, Moment acts like Date given an argument to constructor, and we don't need the overhead.
Was going to go with epoch 0 as the default value, but went with undefined, input on that call appreciated. Epoch 0 lets people parse naturally, but gives a "real" value that isn't true and will throw off analysis without further intervention. Undef will force people to try/catch around Date construction which is inconvenient, especially if you're relying on libraries. Can easily change it to a static epoch 0 string.
Given the data I've seen, we'll have mixed results with this. Integers in the 800,000 range I've seen a few times, which will give us milliseconds after epoch, obviously incorrect (and clearly aren't second-resolution epoch times either). Strings are dealt with via Date.parse (http://tools.ietf.org/html/rfc2822#page-14). If we want to get more sophisticated than that we'll have to start trying format string heuristics via Moment.
Date copy-constructs given an existing Date, I think type-checking vs throwing away a temporary is a toss-up performance wise. Until you start providing custom formats, Moment acts like Date given an argument to constructor, and we don't need the overhead.
Was going to go with epoch 0 as the default value, but went with undefined, input on that call appreciated. Epoch 0 lets people parse naturally, but gives a "real" value that isn't true and will throw off analysis without further intervention. Undef will force people to try/catch around Date construction which is inconvenient, especially if you're relying on libraries. Can easily change it to a static epoch 0 string.
Given the data I've seen, we'll have mixed results with this. Integers in the 800,000 range I've seen a few times, which will give us milliseconds after epoch, obviously incorrect (and clearly aren't second-resolution epoch times either). Strings are dealt with via Date.parse (http://tools.ietf.org/html/rfc2822#page-14). If we want to get more sophisticated than that we'll have to start trying format string heuristics via Moment.
Thoughts?