r****m 发帖数: 70 | 1 自己总结的一些思路,欢迎讨论
1. 需求分析,问问题,列出input, output,可以根据input, output设计API /
interface
2. 分析流程,设计use scenario, 可以用(Given, When, Then)关键词描述
eg: 取款流程
Give a person has a bank account with balance 100
When the person withdraw 30
Then the balance will be changed to 70
3. 根据use scenario设计data model
将上面例子中的名词抽取出来作为对象或属性,动作抽取作为方法
class Person{
long personId;
List accounts;
}
class BankAccount{
long accountId
double balance;
... 阅读全帖 |
|
r****m 发帖数: 70 | 2 9月份的面试,连续四天面了LGTF,准备面试的半年多时间来从本版受益匪浅,现在把
面经写出来回馈本版,希望大家把好的传统延续下去。
L偏重设计,也可能与面的组是platform有关,6个面试有三个是设计,而且涉及很多细
节,比如index,distribute hash, circule counting. 有一面是manager问项目,个
人觉得选一个自己从头到尾做过的项目,然后按我下面的6点进行准备,基本就够了。
L是有题库的,建议多刷版面和glassdoor。
G偏重coding,每一面都是coding开始,而且占很大比例,如果时间多的话可能有两个
coding,也有可能接一个design问题。
T的面试最没规律,感觉基本是面试官自己决定问什么,所以这里不怎么好做总结。
F的面试是最标准化的,两个半coding + 一个design + 半个项目介绍 (项目介绍同上
面L的), F的题目重现率比较高,看版上的题目就差不多了,design问题基本在之前版
上归纳的几个类别: 设计feed,message, search,存储,都和大数据沾边。
LFT面试官大部分是同胞,大部分同胞是... 阅读全帖 |
|
r****m 发帖数: 70 | 3 9月份的面试,连续四天面了LGTF,准备面试的半年多时间来从本版受益匪浅,现在把
面经写出来回馈本版,希望大家把好的传统延续下去。
L偏重设计,也可能与面的组是platform有关,6个面试有三个是设计,而且涉及很多细
节,比如index,distribute hash, circule counting. 有一面是manager问项目,个
人觉得选一个自己从头到尾做过的项目,然后按我下面的6点进行准备,基本就够了。
L是有题库的,建议多刷版面和glassdoor。
G偏重coding,每一面都是coding开始,而且占很大比例,如果时间多的话可能有两个
coding,也有可能接一个design问题。
T的面试最没规律,感觉基本是面试官自己决定问什么,所以这里不怎么好做总结。
F的面试是最标准化的,两个半coding + 一个design + 半个项目介绍 (项目介绍同上
面L的), F的题目重现率比较高,看版上的题目就差不多了,design问题基本在之前版
上归纳的几个类别: 设计feed,message, search,存储,都和大数据沾边。
LFT面试官大部分是同胞,大部分同胞是... 阅读全帖 |
|
B****a 发帖数: 154 | 4 Tables and keys:
patient (patientID, lastName, firstName, bMonth, bDay, bYear,
accountID) key: patientID
account (accountID, balance, responsibleParty) key: accountID
visit (patientID, vMonth, vDay, vYear, description, charge,
doctorID) key: patientID, vMonth, vDay, vYear
doctor (doctorID, drName, degreeYear) key: doctorID
diagnosis (dCode, dDescription) key: dCode
visitDiagnosis (patientID, vMonth, vDay, vYear, dCode) key:
patientID, vMonth, vDay, vYear, dCode
Query: Which patient(s) have been |
|
s******g 发帖数: 287 | 5 比如说设计一个银行的数据库,
有一个table叫transaction(transactionID, accountID, date, ...)
-------------
这个表显然经常需要更新,主要是一条一条的insert操作
这样看来,为了performance,尽量减少index,那么不应该在accountID上建index
不然,每次insert都需要更新这个index(?)
然而,用户(和银行)也需要查询某账户近期的所有transactions
这样似乎又需要给accountID建一个index
想请教一下实际设计中是怎么做的?谢谢! |
|
|
C****g 发帖数: 2220 | 7 Disputes over oily waters: a case study of continental shelf problems and
foreign oil investments in the east china sea and taiwan strait
Subject International law;
International relations
Classification 0616: International law, 0616: International relations
Identifier / keyword Social sciences
Title DISPUTES OVER OILY WATERS: A CASE STUDY OF CONTINENTAL SHELF PROBLEMS
AND FOREIGN OIL INVESTMENTS IN THE EAST CHINA SEA AND TAIWAN STRAIT
Author MA, YING-JEOU
Pages 1
Number of pages 0
Publication y... 阅读全帖 |
|
|
|
|
|
z****u 发帖数: 241 | 12 ---- 2012-02-01 -------
眨眼一月这就过完,美西周三晚上七点.有问有答分享经验,上善若水道法自然!
19:00 - 19:10 校长致辞:土拨鼠抱春,先拿offer后说钱
19:11 - 22:00 找工作与职场问题 问与答
-- 小妹面试,如何说钱?
-- 找工过程,两个阶段的任务和转变点.
-- "这个offer,先问你同意比同意,然后CEO才签,签后就不能改了." 这种说法是真的吗?
-- 背景调查,怎么查,查啥? 如被fire过,能不能被查到?
-- 有统计硕士,如何走找工作捷径,华丽变身?
-- 绿树工作中的问题与回答妹妹关于转行的问题.
-- 知青大哥用直白的话说出找工作秘密.
-- 用复杂的科研头闹想业界的简单问题, 博士就是酱紫滴可爱.
录音如下免费大家听:
freeconferencecalling.com/Recordings/Recording.aspx?fileid=MN3488_
02012012185910998_1050902&bridge=404101&email=&accountid=404101 |
|
t****n 发帖数: 263 | 13 Which database are you using? If it is SQL Server, it looks like accountID+
date is a good candidate for the clustered index. In SQL Server, a clustered
index is actually the table itself, so have no fear of inserting. |
|
j*****n 发帖数: 1781 | 14 that's why there is another thing called fill factor. which can reduce
fragmentation / page split problem.
for me, i'd use combination of transactionID, accountID, and date for the
clustered index. why? most likely transactionID is incremental. it
guarantees new row is appended into the last data page. what about "hot spot
"? it is not a problem for SQL Server 7.0 and later versions if less 1,000
inserts per second. |
|
B*****g 发帖数: 34098 | 15 不明白。how "guarantees new row is appended into the last data page" will
imporve the search based on accountID?
spot |
|
j*****n 发帖数: 1781 | 16 less fragmentation will improve performance, right?
accountid is in the clustered index, index seek will perform. |
|
j*****n 发帖数: 1781 | 17 hmm... you right. however, sql serve can still use B tree instead of hash
tree.
if you doubt with that, we can go for non-clustered index with accountid +
date with fill factor = 80%; then reindex nightly. should work. |
|
c**t 发帖数: 2744 | 18 using profile, I figured what went wrong. I have Account.hbm.xml as
...
The issue was nHibernate always passes in "Account_Id" for my sql-query...
it's not SomeId. |
|
|
p********r 发帖数: 1465 | 20 proc sort... nodupkey out= dupout=;
by accountid;
run;
out= 输出的是不重复的,dupout输出的是所有重复过的。
用compress,之前的童鞋说过了。
million级别的数据也不用太久,很快能出结果。 |
|