SQL Table / Sub-Query Alias Conventions

问题

I've been writing SQL for a number of years now on various DBMS (Oracle, SQL Server, MySQL, Access etc.) and one thing that has always struck me is the seemingly lack of naming convention when it comes to table & sub-query aliases.

I've always read that table alises are the way to go and although I haven't always used them, when I do I'm always stuck between what names to use. I've gone from using descriptive names to single characters such as 't', 's' or 'q' and back again. Take for example this MS Access query I've just written, I'm still not entirely happy with the aliases I'm using even with a relatively simple query as this, I still don't think it's all that easy to read:

SELECT stkTrans.StockName
    , stkTrans.Sedol
    , stkTrans.BookCode
    , SUM(IIF(stkTrans.TransactionType="S", -1 * stkTrans.Units, 0)) AS [Sell Shares]
    , SUM(IIF(stkTrans.TransactionType="B", stkTrans.Units, 0)) AS [Buy Shares]
    , SUM(IIF(stkTrans.TransactionType="B", -1 * stkTrans.Price, 0) * stkTrans1.Min_Units) + SUM(IIF(stkTrans.TransactionType="S", stkTrans.Price, 0) * stkTrans1.Min_Units) AS [PnL]
    , "" AS [Comment]
FROM tblStockTransactions AS stkTrans 
INNER JOIN (SELECT sT1.BookCode
                    , sT1.Sedol
                    , MIN(sT1.Units) AS [Min_Units]
            FROM tblStockTransactions sT1
            GROUP BY sT1.BookCode, sT1.Sedol
            HAVING (SUM(IIF(sT1.TransactionType="S", 1, 0)) > 0
            AND SUM(IIF(sT1.TransactionType="B", 1, 0)) > 0)) AS stkTrans1 ON (stkTrans.BookCode = stkTrans1.BookCode) AND (stkTrans.Sedol = stkTrans1.Sedol)
GROUP BY stkTrans.BookCode, stkTrans.StockName, stkTrans.Sedol;

What do you think? Thought I would throw it out there to see what everyone else's feelings are about this.

回答1:

I don't know of any canonical style rules for naming table/query aliases across databases, although I understand that Oracle recommends abbreviations of three to four characters.

I would generally steer clear of single letter abbreviations, except where the query is sufficiently simple that these should be completely unambiguous to anyone having to maintain the code - typically no more than two or three tables per query.

I would also generally avoid long alias names that conform to the general style of your database table-naming conventions, since it can become unclear what is a database table name and what is an alias.

In the example provided, the alias sT1 inside the inline view is utterly unnecessary, as there is only one table being accessed within that inline view. That leaves one table being joined to one inline view (based on the same table) in the query - in these circumstances, I would use s as the alias for the table, and s1 as the alias for the inline view (to indicate that it was querying the same underlying database table).

回答2:

DT1 and DT2 seems to be good approach .. I am in phase of understanding the existing Procedures and some Procedures use this naming convention DT1,DT2.. It becomes fairly simple to understand ..rather than giving some table short name as alias

回答3:

If I was using SQL Server I'd probably put the derived table in a Common Table Expression (CTE) with a logical name (to indicate what the table is) then shorten it with a correlation name ("table alias") in the main query (to aid readability) e.g.

WITH StockTransactions__type_S_or_B__smallest_units
     AS
     (
      <derived table query here>
     )
SELECT stkTrans.StockName
       ...
FROM tblStockTransactions AS stkTrans 
INNER JOIN StockTransactions__type_S_or_B__smallest_units AS stkTrans1 
   ON (stkTrans.BookCode = stkTrans1.BookCode) AND (stkTrans.Sedol = stkTrans1.Sedol)
GROUP BY stkTrans.BookCode, stkTrans.StockName, stkTrans.Sedol;

Obviously, this isn't an option in Access so you go straight to the correlation name and lose the full name entirely. This is not ideal but acceptable, IMO.

SQL requires a name to be assigned to a derived table for no reason at all. This example from Hugh Darwen, from which I think we can safely assume that the obligation annoys him:

SELECT DISTINCT E#, TOTAL_PAY 
FROM ( SELECT E#, SALARY + BONUS AS TOTAL_PAY
FROM EMP ) AS TEETH_GNASHER
WHERE TOTAL_PAY >= 500

Personally, for such a meaningless requirement I choose the almost meaningless and uncontroversial name, DT1 being a contraction of first derived table and allowing for DT2, DT3, etc e.g.

SELECT DISTINCT E#, TOTAL_PAY 
FROM ( SELECT E#, SALARY + BONUS AS TOTAL_PAY
FROM EMP ) AS DT1
WHERE TOTAL_PAY >= 500

来源：https://stackoverflow.com/questions/9664776/sql-table-sub-query-alias-conventions

标签

sql

coding-style

naming-conventions

alias