In a Java system I'm working on where we simulate trading, we end up with order IDs in the logfiles. Unfortunately, Java's default is to use randomized memory addresses or something for object hash codes, which vary randomly from run to run. This means we can't usefully diff our logfiles. Also, the IDs are 6 hexadecimal bytes.
I think a more useful way to assign order IDs would be with dates: represent the most useful aspects of the date in a short string. In particular, I'm thinking the last digit of the year (since our backtesting simulations cross years), some uniquish ID for the day inside the year (a bit over 8 bits of data, so we can expect a minimum of two characters for this), and then a serial number inside the day.
There's a standard notation for something close to this in the futures markets: ESZ4 is the ES contract with delivery in December 2014, with the letter Z identifying the month, in some sense identifying the fortnight within the year. If we use letters for fortnights in this way, we have 14 or 15 days to discriminate among with an additional letter or number to get to the day, which is entirely doable. Ideally this second letter will be lowercase, and avoid easily misread letters like l, i, and o. Then we can number the orders inside the day with digits.
This gives us IDs like Z4d3, for the third order on 2014-12-04. This is a huge improvement over the current situation: it's a third shorter, human-readable, and consistent from run to run.
From http://www.cmegroup.com/product-codes-listing/month-codes.html:
January F
February G
March H
April J
May K
June M
July N
August Q
September U
October V
November X
December Z
If we wanted to assign two letters per month, sequentially, consistent with the above, we could almost do it:
January E F
February G ???
March H I
April J ???
May K L
June M ???
July N P
August Q R
September S T
October U V
November W X
December Y Z
If we add B for the second half of February, A for the second have of
April, and C for the second half of July, then we can do it, at some
cost to consistency. If we're willing to accept that kind of
inconsistency, then at the cost of a little more of it, we could
eliminate another one: in the above chart, the CME letter is sometimes
the first half of the month and sometimes the second. If we make it
always the first half, and use out-of-sequence letters for the others,
then we end up with this (the six out-of-sequence letters marked with
*):
January F B *
February G C *
March H I
April J D *
May K L
June M E *
July N O
August Q R
September U P *
October V W
November X Y
December Z A *