Validation using Composite Pattern - java

Can somebody please explain how can I use the composite design pattern to validate a registration form in android. In this form I have 8 EditText fields (Ussername, Password, Password re-enter, City, Gender, Age, Phone number, Email). I have already validated it using this code
public boolean validateInput() {
boolean validated = true;
if (mUserValues.getCity() == null || mUserValues.getCity().isEmpty()) {
mCityEditText.setError(getString(R.string.error));
validated = false;
}
if (mUserValues.getEmail() == null || mUserValues.getEmail().isEmpty()) {
Log.d("mEmail", "in mEmail");
mEmail.setError(getString(R.string.error));
validated = false;
}
if (mUserValues.getPassword() == null
|| mUserValues.getPassword().isEmpty()) {
mPasswordEditText.setError(getString(R.string.error));
validated = false;
}
if (mUserValues.getPasswordCheck() == null
|| mUserValues.getPasswordCheck().isEmpty()) {
mPasswordConfirm.setError(getString(R.string.error));
validated = false;
}
if (mUserValues.getPassword() != null
&& mUserValues.getPasswordCheck() != null
&& !mPasswordEditText.getText().toString()
.equals(mPasswordEditText.getText().toString())) {
mPasswordEditText.setError(getString(R.string.pass_check));
mPasswordConfirm.setError(getString(R.string.pass_check));
validated = false;
}
if (mUserValues.getSex() == null || mUserValues.getSex().isEmpty()) {
mSex.setError(getString(R.string.error));
validated = false;
}
if (mUserValues.getPhoneNumber() == 0) {
mPhone.setError(getString(R.string.error));
validated = false;
}
Pattern pattern = Pattern.compile(Constants.EMAIL_PATTERN);
Matcher matcher = pattern.matcher(mEmail.getText().toString());
mEmailCheck = matcher.matches();
if (mEmailCheck == false) {
mEmail.setError(getString(R.string.email_valid));
validated = false;
}
if (mUserValues.getUsername().length() <= 2) {
mUsername.setError(getString(R.string.username_check));
validated = false;
}
if ((mUserValues.getAge() < 18 || mUserValues.getAge() > 150)
&& mUserValues.getAge() > 0) {
mAge.setError(getString(R.string.age_check));
validated = false;
}
return validated;
}
A friend of mine told me that it can be done using the composite design pattern, but I don't really understand how can that be done.
I will be really grateful if somebody could explain.
Thanks in advance.

You are currently inspecting the state of an object (e.g. if (mUserValues.getCity() == null)) from the outside. And if you add a field to mUserValues you'll also have to add code to handle that field to validateInput.
That is bad encapsulation / strong coupling according to the OO principles. That object should handle / validate it's internal state itself (or know another object that can do it -> composite pattern).
I'd move the validation inside the object in a first step (e.g. if (!mUserValues.validateCity())).
After that you'd still need to know about the internals of your mUserValues object on the ouside - e.g. know that there is a city field to validate.
So in a next step you could get rid of that as well. The result could look roughly like
public boolean validateInput() {
boolean validated = true;
for (ValidatableField field : mUserValues.getValidatableFields()) {
if (!field.isValid()) {
field.showError(R.string.error);
validated = false;
}
}
return validated;
}
Now the validateInput code is completely independant of the amount of fields and how they have to be validated. Adding a filed does not require to change anything in there.
Wether you use composition or inheritance is up to you. But using Composition over inheritance is a good thing.

Related

Better condition in java for conditional code [duplicate]

I used spring boot to develop a shell project used to send email, e.g.
sendmail -from foo#bar.com -password foobar -subject "hello world" -to aaa#bbb.com
If the from and password arguments are missing, I use a default sender and password, e.g. noreply#bar.com and 123456.
So if the user passes the from argument they must also pass the password argument and vice versa. That is to say, either both are non-null, or both are null.
How do I check this elegantly?
Now my way is
if ((from != null && password == null) || (from == null && password != null)) {
throw new RuntimeException("from and password either both exist or both not exist");
}
There is a way using the ^ (XOR) operator:
if (from == null ^ password == null) {
// Use RuntimeException if you need to
throw new IllegalArgumentException("message");
}
The if condition will be true if only one variable is null.
But I think usually it's better to use two if conditions with different exception messages. You can't define what went wrong using a single condition.
if ((from == null) && (password != null)) {
throw new IllegalArgumentException("If from is null, password must be null");
}
if ((from != null) && (password == null)) {
throw new IllegalArgumentException("If from is not null, password must not be null");
}
It is more readable and is much easier to understand, and it only takes a little extra typing.
Well, it sounds like you're trying to check whether the "nullity" condition of the two is the same or not. You could use:
if ((from == null) != (password == null))
{
...
}
Or make it more explicit with helper variables:
boolean gotFrom = from != null;
boolean gotPassword = password != null;
if (gotFrom != gotPassword)
{
...
}
Personally, I prefer readable to elegant.
if (from != null && password == null) {
throw new RuntimeException("-from given without -password");
}
if (from == null && password != null) {
throw new RuntimeException("-password given without -from");
}
Put that functionality in a 2 argument method with the signature:
void assertBothNullOrBothNotNull(Object a, Object b) throws RuntimeException
This saves space in the actual method you are interested in and makes it more readable. There is nothing wrong with slightly verbose method names and there is nothing wrong with very short methods.
A Java 8 solution would be to use Objects.isNull(Object), assuming a static import:
if (isNull(from) != isNull(password)) {
throw ...;
}
For Java < 8 (or if you don't like using Objects.isNull()), you can easily write your own isNull() method.
Here is a general solution for any number of null checks
public static int nulls(Object... objs)
{
int n = 0;
for(Object obj : objs) if(obj == null) n++;
return n;
}
public static void main (String[] args) throws java.lang.Exception
{
String a = null;
String b = "";
String c = "Test";
System.out.println (" "+nulls(a,b,c));
}
Uses
// equivalent to (a==null & !(b==null|c==null) | .. | c==null & !(a==null|b==null))
if (nulls(a,b,c) == 1) { .. }
// equivalent to (a==null | b==null | c==null)
if (nulls(a,b,c) >= 1) { .. }
// equivalent to (a!=null | b!=null | c!=null)
if (nulls(a,b,c) < 3) { .. }
// equivalent to (a==null & b==null & c==null)
if (nulls(a,b,c) == 3) { .. }
// equivalent to (a!=null & b!=null & c!=null)
if (nulls(a,b,c) == 0) { .. }
Since you want to do something special (use defaults) when both sender and password are absent, handle that first.
After that, you should have both a sender and a password to send an e-mail; throw an exception if either is missing.
// use defaults if neither is provided
if ((from == null) && (password == null)) {
from = DEFAULT_SENDER;
password = DEFAULT_PASSWORD;
}
// we should have a sender and a password now
if (from == null) {
throw new MissingSenderException();
}
if (password == null) {
throw new MissingPasswordException();
}
An added benefit is that, should either of your defaults be null, that will be detected as well.
Having said that, in general I think that use of XOR should be permissible when that is the operator you need. It is a part of the language, not just some trick that works because of an arcane compiler-bug.
I once had a cow-orker who found the ternary operator too confusing to use...
I would like to suggest another alternative which is how I would actually write this piece of code:
if( from != null )
{
if( password == null )
error( "password required for " + from );
}
else
{
if( password != null )
warn( "the given password will not be used" );
}
To me this seems to be the most natural way to express this condition which makes it easy to understand for somebody who might have to read it in the future. It also allows you to give more helpful diagnostic messages and treat the unnecessary password as less serious and it makes it easy to modify which is rather likely for such a condition. I.e. you may find out that giving a password as a command line argument is not the best idea and may want allow reading the password from standard input optionally if the argument is missing. Or you may want to silently ignore the superfluous password argument. Changes like these would not require you to rewrite the whole thing.
Besides that it executes only the minimum number of comparisons, so it's not more expensive than the more "elegant" alternatives. Although performance is very unlikely a problem here because starting a new process is already much more expensive than a extra null check.
I think a correct way to handle this is to consider three situations: both 'from' and 'password' are provided, neither are provided, a mix of the two are provided.
if(from != null && password != null){
//use the provided values
} else if(from == null && password == null){
//both values are null use the default values
} else{
//throw an exception because the input is not correct.
}
It sounds like the original question wants to break the flow if it is incorrect input, but then they will have to repeat some of the logic later. Perhaps a good throw statement might be:
throw new IllegalArgumentException("form of " + form +
" cannot be used with a "
+ (password==null?"null":"not null") +
" password. Either provide a value for both, or no value for both"
);
Here's a relatively straight-forward way that does not involve any Xor og lengthy ifs. It does however require you to be slightly more verbose, but on the upside, you can use the custom Exceptions I suggested to get a more meaningful error message.
private void validatePasswordExists(Parameters params) {
if (!params.hasKey("password")){
throw new PasswordMissingException("Password missing");
}
}
private void validateFromExists(Parameters params) {
if (!params.hasKey("from")){
throw new FromEmailMissingException("From-email missing");
}
}
private void validateParams(Parameters params) {
if (params.hasKey("from") || params.hasKey("password")){
validateFromExists(params);
validatePasswordExists(params);
}
}
Nobody seems to have mentioned the ternary operator:
if (a==null? b!=null:b==null)
Works nicely for checking this particular condition, but doesn't generalize well past two variables.
As I see your intentions, there is no need to always check both exclusive nullities but to check if password is null if and only if from is not null. You can ignore the given password argument and use your own default if from is null.
Written in pseudo must be like this:
if (from == null) { // form is null, ignore given password here
// use your own defaults
} else if (password == null) { // form is given but password is not
// throw exception
} else { // both arguments are given
// use given arguments
}
I'm surprised nobody mentioned the simple solution of making from and password fields of a class and passing a reference to an instance of that class:
class Account {
final String name, password;
Account(String name, String password) {
this.name = Objects.requireNonNull(name, "name");
this.password = Objects.requireNonNull(password, "password");
}
}
// the code that requires an account
Account from;
// do stuff
Here from could be null or non-null and if it's non-null, both its fields have non-null values.
One advantage of this approach is that the error of making one field but not the other field null gets triggered where the account is initially obtained, not when the code using the account runs. By the time the code using the account is executed, it's impossible for the data to be invalid.
Another advantage to this approach is more readable as it provides more semantic information. Also, it's likely that you require the name and password together in other places so the cost of defining an additional class amortizes over multiple usages.
if ((from == null) == (password == null)){ //if both are true or both are false
}
(Source: Intellij IDEA)

Cleaner way for if-else condition

I have a method having multiple if-else conditions (which is growing with each new msg-type support)
public Message<?> doTransform(Message<String> message) throws TransformationException {
try {
MessageBuilder<String> messageBuilder = null;
String payload = message.getPayload();
String payloadSubStr = payload.substring(0, Math.min(payload.length(), 100));
if(payloadSubStr.contains("<Management>")){
messageBuilder = buildManagementMsg(message);
} else if (payloadSubStr.contains("<Administration>") || (payloadSubStr.contains("<OtherAdministationAlert>"))){
messageBuilder = buildAdminMessages(message);
} else if (payloadSubStr.startsWith("Council")){
messageBuilder = parseCouncilMessages(message);
} else if (payloadSubStr.indexOf("Security") >= 0
|| payloadSubStr.indexOf("OtherSecurityAlert") >= 0){
messageBuilder = buildSecurityMessages(message);
} else if ( payloadSubStr.indexOf("<Staff>") >= 0
|| payloadSubStr.indexOf("<OtherStaffAlert>") >= 0){
messageBuilder = buildStaffMessages(message);
} else if(payloadSubStr.indexOf("<Student>") >= 0) {
messageBuilder = buildStudentMessages(message);
}else {
messageBuilder = buildOtherMessages(message);
}
return messageBuilder.build();
} catch(Exception e) {
throw new TransformationException(e);
}
}
Reason for doing substring: To avoid complete msg traversing with each if/else condition
Reason for using contains/indexOf combination: Messages received by this method can vary
Wanted to replace these if/else statements with some more cleaner logic. Not getting if Switch/Enum can be used or need to use any pattern as suggested over https://www.baeldung.com/java-replace-if-statements.
Gone through various similar questions available but not getting anything. Any suggestion will be helpful.
Thanks
If you are concerned with the number of conditions you'd need to add you could introduce a list of "message builder factories" (for lack of a better name atm) that you could append to.
A factory would contain a predicate to test the substring for and a messageBuilder(...) method. Then you'd iterate over the factory, check each of the predicates and if execute messageBuilder() on the first that matches.
Example:
interface MessageBuilderFactory<T> {
boolean test(T payload);
MessageBuilder<T> messageBuilder(Message<T> message);
}
class ManagementMBFactory implements MessageBuilderFactory<String> {
boolean test(String payload) {
return payload.contains("Management");
}
MessageBuilder<String> messageBuilder(Message<String> message) {
//content of buildManagementMsg() here
}
}
And in your code:
List<MessageBuilderFactory<String>> factories = ... //get the list of factories from somewhere
for( MessageBuilderFactory<String> factory : factories) {
if( factory.test(payloadSubStr) {
messageBuilder = factory.messageBuilder(message);
}
}
An advantage of doing it that way would be that the list of possible message builder factories is easily available and classes can be kept small (not all those buildXxx() methods in one single class).
Alternatively, if your message payload allows for that, you could actually try to parse it (it looks like XML) and operate on events, i.e. elements being found. That might be faster in the case of many small payloads and a huge number of possible message builders.
People are divided on the idea of multiple return statements in java but in this case I think I would tend to do that:
if(payloadSubStr.contains("<Management>")){
return buildManagementMsg(message);
}
if (payloadSubStr.contains("<Administration>") || (payloadSubStr.contains("<OtherAdministationAlert>"))) {
return buildAdminMessages(message);
}
if (payloadSubStr.startsWith("Council")){
return parseCouncilMessages(message);
}
if (payloadSubStr.indexOf("Security") >= 0
|| payloadSubStr.indexOf("OtherSecurityAlert") >= 0){
return buildSecurityMessages(message);
}
if ( payloadSubStr.indexOf("<Staff>") >= 0
|| payloadSubStr.indexOf("<OtherStaffAlert>") >= 0){
return buildStaffMessages(message);
}
if (payloadSubStr.indexOf("<Student>") >= 0) {
return buildStudentMessages(message);
}
return buildOtherMessages(message);
Going one step further this could be done with a validation service.
public class ValidationService {
public boolean isManagement(String str) { return str.contains("<Management>");
// ... and so on
}
And you can inject the service into the code such that you have
if (validationService.isManagement(payloadSubStr)) {
return buildManagementMsg(message);
}
// ...
For a case with conditions currently in OR you could use a list in the service, for example
public boolean isSecurity(String str) {
for (String term : new String[]{"Security", "OtherSecurityAlert"}) {
if (str.contains(term)) {
return true;
}
}
return false;
}

SonarQube Refactor this method to reduce its Cognitive Complexity

I have the below utility method and I am using multiple if statements and getting cognitive complexity issue. I went through some links, but I am not able to understand how should I change my code without affecting users of this method.
public static boolean isWrapperValid(WrapperClass wrapper, boolean isTechnicalToken){
String key=null;
boolean isValidWrapper = false;
if (wrapper != null && wrapper.length() > 7
&& wrapper.substring(0, 6).equalsIgnoreCase("XYZ"))
{
wrapper= wrapper.substring(7, wrapper.lastIndexOf('.')+1);
}
if(wrapper != null && wrapper.equalsIgnoreCase("TFR")) {
isValidWrapper=Boolean.TRUE;
}
try {
key = wrapper.getKey();
}
catch (Exception exception) {
return isValidWrapper;
}
if(key!=null) {
Date tokenExpiryTime = key.getExpiresAt();
if(tokenExpiryTime!=null) {
return isValidWrapper;
}
String algorithm=key.getAlgorithm();
if(!DESIRED_ALGO.equals(algorithm)) {
return isValidWrapper;
}
String value6=key.getType();
if(!DESIRED_TYPE.equals(value6)) {
return isValidWrapper;
}
if(key.getValue1()!=null && key.getValue2().size()>0 && key.getValue3()!=null && key.getValue4()!=null && key.getValue5()!=null) {
isValidWrapper=Boolean.TRUE;
}
}
return isValidWrapper;
}
Please share your suggestions to refactor this code.
I don't think that merging many if conditions to one or simply do a code clean up, for example by changing the order of some instructions, can solve your problem.
Your code does not match the single responsibility principle. You should refactor this big method to smaller parts. Due to this it will testable, easier to maintain and read. I spent some time and did this:
public static boolean isWrapperValid(WrapperClass wrapper, boolean isTechnicalToken) {
final WrapperClass unpackedWrapper = unpackWrapper(wrapper);
boolean wrapperValid = isUnpackedWrapperValid(unpackedWrapper);
Key key = null;
try {
key = unpackedWrapper.getKey();
} catch (final Exception exception) {
return wrapperValid;
}
if (key != null) {
if (doesKeyMeetsBasicConditions(key)) {
return wrapperValid;
}
if (doesKeyMeetsValueConditions(key)) {
return true;
}
}
return wrapperValid;
}
protected static WrapperClass unpackWrapper(final WrapperClass wrapper) {
if (wrapper != null && wrapper.length() > 7 && wrapper.substring(0, 6).equalsIgnoreCase("XYZ")) {
return wrapper.substring(7, wrapper.lastIndexOf('.') + 1);
}
return wrapper;
}
protected static boolean isUnpackedWrapperValid(final WrapperClass wrapper) {
return wrapper != null && wrapper.equalsIgnoreCase("TFR");
}
protected static boolean doesKeyMeetsBasicConditions(final Key key) {
Date tokenExpiryTime = key.getExpiresAt();
if (tokenExpiryTime != null) {
return true;
}
String algorithm = key.getAlgorithm();
if (!DESIRED_ALGO.equals(algorithm)) {
return true;
}
String value6 = key.getType();
return !DESIRED_TYPE.equals(value6);
}
protected static boolean doesKeyMeetsValueConditions(final Key key) {
return key.getValue1() != null && key.getValue2().size() > 0
&& key.getValue3() != null && key.getValue4() != null
&& key.getValue5() != null;
}
I don't know the domain logic, so some of my methods have stupid names etc. As you can see, now you have a lot of smaller methods with not many branches (if conditions) - easier to test (a static code is not nice, but you can mock it by using for example PowerMock).
A bit of rewriting delivered a simplification, that still could be improved upon.
public static boolean isWrapperValid(WrapperClass wrapper, boolean isTechnicalToken){
if (wrapper != null && wrapper.length() > 7
&& wrapper.substring(0, 6).equalsIgnoreCase("XYZ"))
{
wrapper = wrapper.substring(7, wrapper.lastIndexOf('.')+1);
}
boolean isValidWrapper = wrapper != null && wrapper.equalsIgnoreCase("TFR");
try {
String key = wrapper.getKey();
if (key != null && key.getExpiresAt() == null
&& DESIRED_ALGO.equals(key.getAlgorithm())
&& DESIRED_TYPE.equals(key.getType())
&& key.getValue1() != null && !key.getValue2().isEmpty()
&& key.getValue3() != null && key.getValue4() != null
&& key.getValue5() != null) {
isValidWrapper = true;
}
}
catch (Exception exception) {
// DO NOTHING
}
return isValidWrapper;
}
After comment: here I catch any exception for all calls.
First of all, Sonar should give you more flags: reusing the wrapper parameter is usually a bad practice, NPE where invoking wrapper.getKey because wrapper can be null, but anyway, not the point...
Try reducing the number of if statements by creating local boolean variables (or possibly 1 big if statement if you have less than 5 or 6 tests, but often less readable). Once it's done, you should only have 1 block testing these boolean variables, and have one return statement, like the example above (not necessarily accurate!):
boolean expired = tokenExpiryTime != null;
boolean desiredAlgo = DESIRED_ALGO.equals(key.getAlgorithm());
boolean desiredType = DESIRED_TYPE.equals(value6);
if (expired || !desiredAlgo || !desiredType) {
return isValidWrapper;
}
However, your Cognitive complexity level seems pretty low if this kind of algorithm triggers it...
Another big way to reduce an algorithm complexity is to turn sub-blocks of code (loops, if and try-catch) into private methods. In your example, it could be something like a checkWrapperValidity method, responsible for every test returning isValidWrapper

SonarQube - boolean logic correctness -

I have a problem with the logic expression on my method matches1().
Problem
SonarQube is telling me there is an error:
(expectedGlobalRule == null && actual != null)
SonarQube:
Change this condition so that it does not always evaluate to
"true".
Conditions should not unconditionally evaluate to "TRUE" or to "FALSE"
I'm essentially doing this logic to avoid a NPE on my "Block to be executed".
My code
matches1()
private boolean matches1(GbRule actual, GbRule expected) {
if(actual == null && expected == null) {
return true;
} else if((expected == null && actual != null) || (expected != null && actual == null)) {
return false;
} else {
//Block to be executed
}
}
I inverted the logic in to see what SonarQube would tell me and he doesn't complain about it.
matches2()
private boolean matches2(GbRule actual, GbRule expected) {
if(actual == null && expected == null) {
return true;
} else if(expected != null && actual != null) {
//Block to be executed
} else {
return false;
}
}
Question
Do the problem is in my boolean logic or it's SonarQube that lost
his mind?
If the problem is within sonarQube, how could I resolve it?
The problem is in your logic. Let's take it piece by piece:
if(actual == null && expected == null) {
return true;
At this point if both vars are null then we're no longer in the method. So if we get any further, then at least one of them is non-null.
The viable options at this point are:
actual = null, expected = non-null
actual = non-null, expected = null
actual = non-null, expected = non-null
Now, let's look at the next bit of code:
} else if((expected == null && actual != null)
We already know that both variables can't be null, so as soon as we know expected == null, there's no need to test whether actual != null. That has already been proven by the fact that we got this far. So actual != null is always true, which is why an issue is raised.
Edit
This means that your code could be boiled down to:
private boolean matches1(GbRule actual, GbRule expected) {
if(actual == null && expected == null) {
return true;
} else if(actual == null || expected == null) {
return false;
}
//Block to be executed
}
Note that the else isn't needed & dropping it makes the code easier to read.
Even when the code is correct; seriously, it makes my eyes hurt. Thing is: it is hard to read. Such kind of nested conditions is something that one should not be writing in the first place.
If you can't avoid it; at least refactor it into something like
private boolean areActualAnedExpectedBothNull(args ...) {
return actual == null && expectedGlobalRule == null;
}
And please note; you can dramatically simply your code:
if (areActualAnedExpectedBothNull(actual, expected)) {
return true;
}
if (actual == null) {
return false;
}
if (expected == null) {
return false;
}
do your thing ...
and use such methods in your other code. And of course, you do a lot of unit testing; probably with coverage measurements; just to make sure that your tests really test all possible paths through this maze.
But as said; you better step back and think if there are ways to avoid writing such code in the first place.
The typical answer to booleans, and if/else chains in OO programming is polymorphism. So instead of asking something about its state; you turn to interfaces/abstract classes; and have different implementations of those. And then you have a factory giving you that implementation you need; and then you just call methods on that; without further need for if/else/whatever.
If you don't know what I am talking about - watch these videos; especially the second one!
The problem is with SonarQube.
See this article for more info on ignoring that issue: https://www.bsi-software.com/en/scout-blog/article/ignore-issues-on-multiple-criteria-in-sonarqube.html
You can just set it up to ignore that error within that file.
The gist of it is
Open the Settings (SonarQube General Settings or project Settings) and
select the Exclusions category. Switch to the Issues Exclusions and
scroll down to “Ignore Issues on Multiple Criteria”. Set squid:S00112
as Rule Key Pattern and **/*Activator.java as File Path Pattern.
You will need to change the rule key pattern to the pattern associated with the rule that is being violated for your code and the file pattern as the path of your .java file.

Check two arguments in Java, either both not null or both null elegantly

I used spring boot to develop a shell project used to send email, e.g.
sendmail -from foo#bar.com -password foobar -subject "hello world" -to aaa#bbb.com
If the from and password arguments are missing, I use a default sender and password, e.g. noreply#bar.com and 123456.
So if the user passes the from argument they must also pass the password argument and vice versa. That is to say, either both are non-null, or both are null.
How do I check this elegantly?
Now my way is
if ((from != null && password == null) || (from == null && password != null)) {
throw new RuntimeException("from and password either both exist or both not exist");
}
There is a way using the ^ (XOR) operator:
if (from == null ^ password == null) {
// Use RuntimeException if you need to
throw new IllegalArgumentException("message");
}
The if condition will be true if only one variable is null.
But I think usually it's better to use two if conditions with different exception messages. You can't define what went wrong using a single condition.
if ((from == null) && (password != null)) {
throw new IllegalArgumentException("If from is null, password must be null");
}
if ((from != null) && (password == null)) {
throw new IllegalArgumentException("If from is not null, password must not be null");
}
It is more readable and is much easier to understand, and it only takes a little extra typing.
Well, it sounds like you're trying to check whether the "nullity" condition of the two is the same or not. You could use:
if ((from == null) != (password == null))
{
...
}
Or make it more explicit with helper variables:
boolean gotFrom = from != null;
boolean gotPassword = password != null;
if (gotFrom != gotPassword)
{
...
}
Personally, I prefer readable to elegant.
if (from != null && password == null) {
throw new RuntimeException("-from given without -password");
}
if (from == null && password != null) {
throw new RuntimeException("-password given without -from");
}
Put that functionality in a 2 argument method with the signature:
void assertBothNullOrBothNotNull(Object a, Object b) throws RuntimeException
This saves space in the actual method you are interested in and makes it more readable. There is nothing wrong with slightly verbose method names and there is nothing wrong with very short methods.
A Java 8 solution would be to use Objects.isNull(Object), assuming a static import:
if (isNull(from) != isNull(password)) {
throw ...;
}
For Java < 8 (or if you don't like using Objects.isNull()), you can easily write your own isNull() method.
Here is a general solution for any number of null checks
public static int nulls(Object... objs)
{
int n = 0;
for(Object obj : objs) if(obj == null) n++;
return n;
}
public static void main (String[] args) throws java.lang.Exception
{
String a = null;
String b = "";
String c = "Test";
System.out.println (" "+nulls(a,b,c));
}
Uses
// equivalent to (a==null & !(b==null|c==null) | .. | c==null & !(a==null|b==null))
if (nulls(a,b,c) == 1) { .. }
// equivalent to (a==null | b==null | c==null)
if (nulls(a,b,c) >= 1) { .. }
// equivalent to (a!=null | b!=null | c!=null)
if (nulls(a,b,c) < 3) { .. }
// equivalent to (a==null & b==null & c==null)
if (nulls(a,b,c) == 3) { .. }
// equivalent to (a!=null & b!=null & c!=null)
if (nulls(a,b,c) == 0) { .. }
Since you want to do something special (use defaults) when both sender and password are absent, handle that first.
After that, you should have both a sender and a password to send an e-mail; throw an exception if either is missing.
// use defaults if neither is provided
if ((from == null) && (password == null)) {
from = DEFAULT_SENDER;
password = DEFAULT_PASSWORD;
}
// we should have a sender and a password now
if (from == null) {
throw new MissingSenderException();
}
if (password == null) {
throw new MissingPasswordException();
}
An added benefit is that, should either of your defaults be null, that will be detected as well.
Having said that, in general I think that use of XOR should be permissible when that is the operator you need. It is a part of the language, not just some trick that works because of an arcane compiler-bug.
I once had a cow-orker who found the ternary operator too confusing to use...
I would like to suggest another alternative which is how I would actually write this piece of code:
if( from != null )
{
if( password == null )
error( "password required for " + from );
}
else
{
if( password != null )
warn( "the given password will not be used" );
}
To me this seems to be the most natural way to express this condition which makes it easy to understand for somebody who might have to read it in the future. It also allows you to give more helpful diagnostic messages and treat the unnecessary password as less serious and it makes it easy to modify which is rather likely for such a condition. I.e. you may find out that giving a password as a command line argument is not the best idea and may want allow reading the password from standard input optionally if the argument is missing. Or you may want to silently ignore the superfluous password argument. Changes like these would not require you to rewrite the whole thing.
Besides that it executes only the minimum number of comparisons, so it's not more expensive than the more "elegant" alternatives. Although performance is very unlikely a problem here because starting a new process is already much more expensive than a extra null check.
I think a correct way to handle this is to consider three situations: both 'from' and 'password' are provided, neither are provided, a mix of the two are provided.
if(from != null && password != null){
//use the provided values
} else if(from == null && password == null){
//both values are null use the default values
} else{
//throw an exception because the input is not correct.
}
It sounds like the original question wants to break the flow if it is incorrect input, but then they will have to repeat some of the logic later. Perhaps a good throw statement might be:
throw new IllegalArgumentException("form of " + form +
" cannot be used with a "
+ (password==null?"null":"not null") +
" password. Either provide a value for both, or no value for both"
);
Here's a relatively straight-forward way that does not involve any Xor og lengthy ifs. It does however require you to be slightly more verbose, but on the upside, you can use the custom Exceptions I suggested to get a more meaningful error message.
private void validatePasswordExists(Parameters params) {
if (!params.hasKey("password")){
throw new PasswordMissingException("Password missing");
}
}
private void validateFromExists(Parameters params) {
if (!params.hasKey("from")){
throw new FromEmailMissingException("From-email missing");
}
}
private void validateParams(Parameters params) {
if (params.hasKey("from") || params.hasKey("password")){
validateFromExists(params);
validatePasswordExists(params);
}
}
Nobody seems to have mentioned the ternary operator:
if (a==null? b!=null:b==null)
Works nicely for checking this particular condition, but doesn't generalize well past two variables.
As I see your intentions, there is no need to always check both exclusive nullities but to check if password is null if and only if from is not null. You can ignore the given password argument and use your own default if from is null.
Written in pseudo must be like this:
if (from == null) { // form is null, ignore given password here
// use your own defaults
} else if (password == null) { // form is given but password is not
// throw exception
} else { // both arguments are given
// use given arguments
}
I'm surprised nobody mentioned the simple solution of making from and password fields of a class and passing a reference to an instance of that class:
class Account {
final String name, password;
Account(String name, String password) {
this.name = Objects.requireNonNull(name, "name");
this.password = Objects.requireNonNull(password, "password");
}
}
// the code that requires an account
Account from;
// do stuff
Here from could be null or non-null and if it's non-null, both its fields have non-null values.
One advantage of this approach is that the error of making one field but not the other field null gets triggered where the account is initially obtained, not when the code using the account runs. By the time the code using the account is executed, it's impossible for the data to be invalid.
Another advantage to this approach is more readable as it provides more semantic information. Also, it's likely that you require the name and password together in other places so the cost of defining an additional class amortizes over multiple usages.
if ((from == null) == (password == null)){ //if both are true or both are false
}
(Source: Intellij IDEA)

Categories