p****1
发帖数: 275 | 1 【 以下文字转载自 JobHunting 讨论区 】
发信人: peace1 (peace1), 信区: JobHunting
标 题: SQL, recruiter发过来的面试题
发信站: BBS 未名空间站 (Thu Sep 8 22:14:20 2016, 美东)
有兴趣做不? 我琢磨着, 出题的人是在寻找最优解 :)
We work with large datasets, and are always performance conscious since
extended processing times will impact our time to market. Keep this in mind
as you answer the following questions:
There is a table defined as:
CREATE TABLE [Positions](
[load_id] [int] NOT NULL,
[acct_cd] [varchar](20) NOT NULL,
[acct_num] [varchar](255) NULL,
[sec_id] [varchar](50) NOT NULL,
[long_sht_cd] [varchar](3) NOT NULL,
[sedol] [varchar](15) NULL,
[isin] [varchar](15) NULL,
[cusip] [varchar](9) NULL,
[sec_type] [varchar](8) NULL,
[sec_name] [varchar](100) NULL,
[currency_cd] [varchar](3) NULL,
[total_holding] [decimal](18, 4) NULL,
[mkt_price] [float] NULL,
[datetime_stamp] [datetime] NULL,
CONSTRAINT [pk_Positions] PRIMARY KEY CLUSTERED (
[load_id] ASC,
[acct_cd] ASC,
[sec_id] ASC,
[long_sht_cd] ASC )
)
This table holds account positions data that are appended to multiple times
a day
There are currently some 24 million rows in the table. Every time we append
additional positions we add approximately 32,000 entries to this table, and
all 32,000 entries will have the same load_id. The load_id is incremented by
one each time we load a batch of 32,000 entries (i.e. the first 32K entries
have load_id=1, the next 32K has load_id=2, etc...). The datetime_stamp
field shows the time at which the entries were loaded and is the same for
all 32K entries in a single load.
How would you efficiently retrieve the first set of positions for the
current day given the above table definition?
Example:
Today, positions were loaded into this table at 8am, 10am and 3pm. At 5pm
today we want to know what positions were loaded at 8am since that is the
first load that occurred today. Note that for any given day, there can be
different number of loads and the times that the loads occur will vary. |
|
