SQL insert unix timestamp with milliseconds - date

Question

Can unix timestamps contain the milliseconds value of the time?
If so, using SQL Server 2008 R2:
CREATE TABLE [dbo].[my_table_calls_log]
(
[id] [bigint] IDENTITY(1,1) NOT NULL,
[requestdate] [bigint] NULL,
[partycode] [bigint] NULL,
CONSTRAINT [PK_my_table_calls_log]
PRIMARY KEY CLUSTERED ([id] ASC)
)
The following select gives me the current date and time with milliseconds:
SELECT CONVERT(VARCHAR,SYSDATETIME(),121);
Example:
2015-09-11 13:29:02.8239061
How do I convert that long YYYY-MM-DD HH:MM:SS.MS date/time value into a unix timestamp so that I can insert it using something like:
DECLARE #UnixDate AS bigint;
SET #UnixDate = "the unix timestamp equivalent of SELECT CONVERT(VARCHAR,SYSDATETIME(),121);"
INSERT INTO my_table_calls_log (requestdate,partycode)
VALUES (#UnixDate,123);

Answer 1

Standard unix timestamps are 32-bit signed integers with a precision of 1 second (see here), that said there is certainly nothing invalid about using a 64-bit integer and using a precision of milliseconds if you so choose (so long as you are certain that all code that will use/consume that timestamp value is aware of that constraint). Most standard unix timestamp libraries assume it is a 1 second precision, and if you pass a 64-bit signed integer instead of a 32-bit integer that still holds and the assumption is the user is just planning for a longer life of the timestamp (i.e. to track date/time values past January 19, 2038, which is the date a 32-bit based unix timestamp will stop working at some point from the default epoch due to an overflow).
Unix timestamps also generally assume a base epoch of '1970-01-01 00:00:00', so using that assumption you can convert a standard SQL Server datetime value to a unixtimestamp using something like this:
select datediff(second, '1970-01-01T00:00:00.000', sysdatetime());
If you want that to be millisecond based, it becomes a bit more complex as the datediff function in SQL Server is 32-bit integer based (i.e. an int data type) and the difference between the default epoch and the current date/time in milliseconds exceeds the upper-bound of that type, so we have to get a little creative, here is one option (which I tend to use wrapped in a function):
declare #start datetime2 = '1970-01-01T00:00:00.000',
#end datetime2 = sysdatetime(),
#ms_in_day bigint = 60 * 60 * 24 * 1000;
select (#ms_in_day * datediff(day, #start, #end)) - datediff(millisecond, #end, cast(#end as date));
In each of these examples, if you have a varchar/character based date/time value, simply insert directly into the script in place of the sysdatetime() function I used.