Sunday, July 13, 2014

Exception - How to manage different exceptions your code wants to throw?

Often I have seen multiple Exception types created to handle different types of exceptions. In Java they all become individual classes. They might all extend from a common Exception class defined by you. But as Java does not support inheritance of constructors we end up having multiple classes with similar code. Not if it was Ruby code :)

Anyways. I got the following thought from the way Http defines Status Codes.

http://www.w3.org/Protocols/rfc2616/rfc2616-sec10.html

Define broad level Exception classes.

Informational, Success, Bad Request, Server Error. You may or may not have Redirection. In each of these Exception classes have status code. If possible, try to keep the status codes matching to the http status codes.

Now you have limited set of classes and you can set specific status code based on a scenario.

class InformationalException {
  public static enum InformationalStatus {
     // you have avoid it by using directly http status from commons client
  }

  public InformationalException(InformationalStatus status, String message, Throwable th) {
  }
}

The good news is most of the clients understand the http codes and it is no longer a magic.

Less code is less pain.

Thursday, June 12, 2014

Oracle Database - reading a CLOB through sqlplus

If you execute  through sqlplus 'select <clob column>' you will get truncated data. To see the complete data in sqlplus

set buffer <X>

select DBMS_LOB.substr(column, <buffer size>) from table

Note: This is required for sqlplus. JDBC and hibernate handle it directly so this 'select column' should be good enough.

Wednesday, June 11, 2014

SQL equal or not equal and Mr NULL

I created a table with a column

IS_CONFIDENTIAL CHAR(1)

and I added this clause to my query if the user should not see the confidential data

AND IS_CONFIDENTIAL <> 'Y'

Now the user could not see the confidential records. All good!

But he could not see few other records which were not confidential ?? I checked the table and found records with IS_CONFIDENTIAL values 'Y','N' and NULL and the records with NULL value didn't show up !!

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

IS_CONFIDENTIAL <> 'Y' does not mean it to return rows with NULL value !!

This is what I did (though there could be multiple solutions and may be better :))

IS_CONFIDENTIAL NOT NULL  CHAR(1)

Or I should have changed my query to AND ( IS_CONFIDENTIAL = 'N' OR IS_CONFIDENTIAL IS NULL)

Note: I agree that positive check is always better. But I thought what if there are more values than Y and N, IS_CONFIDENTIAL<> 'Y' was the choice for me.

Bulb: Did you know you can enable hibernate to print the queries it executes! It is very helpful.

Bulb: CHAR vs VARCHAR


curl command to error out on HTTP error codes

$ curl -I -f "http://json.org/example"
HTTP/1.1 200 OK
Date: Tue, 10 Jun 2014 20:01:02 GMT
Server: Apache
...

Ah! but it does not provide the / at the tail. We get a 404.

$ curl -I "http://json.org/example/"
HTTP/1.1 404 Not Found
Date: Tue, 10 Jun 2014 20:01:15 GMT
Server: Apache
...

echo $? will return 0 in both the case.

What if we are using it inside a script and want the process to fail on non success HTTP status?

Use –f option

$ curl -I -f "http://json.org/example/"
curl: (22) The requested URL returned error: 404

$ echo $?
22

The man says:

       -f, --fail
              (HTTP)  Fail  silently (no output at all) on server errors. This is mostly done to better enable scripts etc to better deal with failed attempts. In
              normal cases when a HTTP server fails to deliver a document, it returns an HTML document stating so (which often also describes why and more).  This
              flag will prevent curl from outputting that and return error 22.

              This  method  is  not  fail-safe  and  there  are occasions where non-successful response codes will slip through, especially when authentication is
              involved (response codes 401 and 407).

Note: -I is better option to user than --request HEAD, as --request HEAD request will hang for a while :)

Tuesday, May 20, 2014

session_privs and session_roles

So my friend was getting 'SYS.DBA_IND_COLUMNS table does not exit' while he was trying to run

desc dba_ind_coulmns

I was able to run the same query. So of course we are in the territory of authorizations. But how do I prove it?

Then can session_privs and session_roles into picture. For me the result was

SQL> select * from session_privs;

PRIVILEGE
----------------------------------------
CREATE SESSION
ALTER SESSION
SELECT ANY TABLE
SELECT ANY SEQUENCE

SQL> select * from session_roles;

ROLE
------------------------------
CONNECT
SELECT_CATALOG_ROLE
HS_ADMIN_ROLE
HS_ADMIN_SELECT_ROLE
HS_ADMIN_EXECUTE_ROLE
ADHOC

6 rows selected.

His user has more privilege but less roles.

For him the role was ADHOC_DML and no admin table roles and hence it was proved why he got 'table does not exist' error.

Another interesting table to know is v$session.

select SID,MACHINE,USERNAME from V$SESSION;

Friday, May 9, 2014

Git commands

> Log commits for a user since. Print in one line and with abbreviated commit

git log --committer=appandey --no-merges --since=2014-04-24 --pretty=oneline --abbrev-commit

> add all java files

git add \*.java

> Commit on someone else's behalf (same team)

git commit --author='author' -m '..'

fatal: No existing author found with 'author'

git commit --author='author <author@company.com>' -m '..'

e.g. git commit --author="amodpandey <amodpandey@gmail.com>" -m '...'

Monday, May 5, 2014

Did you know java TimeUnit class

I am not sure why Date din't do it, Calendar didn't do it and Apache DateUtils didn't do it?

Though every other product in every other company would need date diff !!!

I was introduced to TimeUnit through a retry wrapper class where I had to specify retry time using TimeUnit. So no longer we need a variable like private static final long numberOfSecondsInMinutes  = 60; :)

Date diff is a two liner with this class

1. Get the diff in millis
long diff = date1.getTime() - date2.getTime();

2. TimeUnit.MILLISECONDS.toDays(diff);

nJoy

Friday, April 18, 2014

Http Accept - 406

There are many HTTP headers which go unnoticed. Accept is one such header.

Accept is a way to tell the server what the client accepts. A server can smartly handle this.

e.g. Accept: application/xml to a service returns xml and Accept: application/json to same end point should return json.

Sometimes we have the server side code as to strictly say we support requests with Accept: application/json. I have seen ROR doing it when the return is of type JSON.

If the Web server detects that the data it wants to return is not acceptable to the client, it returns a header containing the 406 error code. 

I have observed the below behavior with our ROR set up

Success

Accept: application/xml,application/json
Accept: application/json,application/xml"
Accept: application/xml;q=0.9,application/json
Accept: */*

Failure (406)

Accept: */*, application/xml;
Accept: application/xml,*/*;
Accept: application/xml;q=0.9,*/*;
Accept: application/xml;

Note: I see */* is successful if Accept has only one entry with */*, else it expects application/json somewhere.

Chrome browser by default sends (may depend on browser and version)

  1. Accept:
    text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8

So again it will fail. There are browser add ons which can help manipulate any of the HTTP headers for calls made through the browser. I have to use that for this call.

HTH someone, someday

Sunday, March 2, 2014

The Orthodox: Java getter and setter

Let me start by saying, 'I hate it'. I always wrote it for the sake of writing it. Before making any decision I would like to bounce it with the community.

In the recent years most of us write Java classes that fall into one of these categories.

1. The so called POJO - Only attributes with getters and setters. Mostly used for serialization/de-serialization (to DB row, to JSON, to XML or any thing). - The carriers

2. The business processing classes where we tend not to have any instance variables. These are like the singleton instances doing the processing accepting/returning some POJO. - The performers

e.g.

class Customer {

public Integer addCustomer(MyCustomerBean bean) {
}

public MyCustomerBean get(Integer custid) {
}
}

So there is no instance variable dependency between two methods.

Further we get details from multiple such POJOs (e.g. MyAccountBean, MyCustomerBean) and put it into another POJO that is response to a service or UI call. These days annotation based mappers are quite famous, which would give you ResponseVO from  MyCustomerBean and MyAccountBean with a single line of code. I do remember how much tough life was without them (e.g. Castor for XML)

In none of these classes, ResponseVO, MyCustomerBean or MyAccountBean, the getters and setters do anything other than getting the variable and setting the variable. Is there any reason for which we keep doing it and blame it on to encapsulation??

Give me any good reason they are required for and yet they occupy 30-40% of your code! Thanks to eclipse to provide us with getters and setters generators. Just to add to the misery we add javadocs to such code ( again thanks to the auto javadoc generators).

Let's think whether they are required for our POJOs' or not? Can we have our POJO attributes public and lessen the pain.

Do I suggest having the Customer class with all the attributes and associated methods? Probably not. Today MyCustomerBean might a hibernate dependent bean and tomorrow it may be Toplink based. Let's give getters and setters a peaceful send off.

It may sound touche but YAGNI principle is really important to follow when the life of a code is too small to think about improvements, mostly it is re-writes.

What if I need to add a logic in my getters and setters won't help! :)

Code less that makes sense!
Happy Coding!