- Programmatic Configuration
- INI Configuration
- Creating a SecurityManager from INI
- INI Sections
- Lend a hand with documentation
Shiro is designed to work in any environment, from simple command-line applications to the largest enterprise clustered applications. Because of this diversity of environments, there are a number of configuration mechanisms that are suitable for configuration. This section covers the configuration mechanisms that are supported by Shiro core only.
|Many Configuration Options|
Shiro's SecurityManager implementations and all supporting components are all JavaBeans compatible. This allows Shiro to be configured with practically any configuration format such as regular Java, XML (Spring, JBoss, Guice, etc), YAML, JSON, Groovy Builder markup, and more.
The absolute simplest way to create a SecurityManager and make it available to the application is to create an org.apache.shiro.mgt.DefaultSecurityManager and wire it up in code. For example:
Surprisingly, after only 3 lines of code, you now have a fully functional Shiro environment suitable for many applications. How easy was that!?
As discussed in the Architecture chapter, Shiro's SecurityManager implementations are essentially a modular object graph of nested security-specific components. Because they are also JavaBeans-compatible, you can call any of the nested components getter and setter methods to configure the SecurityManager and its internal object graph.
For example, if you wanted to configure the SecurityManager instance to use a custom SessionDAO to customize Session Management, you could set the SessionDAO directly with the nested SessionManager's setSessionDAO method:
Using direct method invocations, you can configure any part of the SecurityManager's object graph.
But, as simple as programmatic customization is, it does not represent the ideal configuration for most real world applications. There are a few reasons why programmatic configuration may not be suitable for your application:
- It requires you to know about and instantiate a direct implementation. It would be nicer if you didn't have to know about concrete implementations and where to find them.
- Because of Java's type-safe nature, you're required to cast objects obtained via get* methods to their specific implementation. So much casting is ugly, verbose, and tightly-couples you to implementation classes.
- The SecurityUtils.setSecurityManager method call makes the instantiated SecurityManager instance a VM static singleton, which, while fine for many applications, would cause problems if more than one Shiro-enabled application was running on the same JVM. It could be better if the instance was an application singleton, but not a static memory reference.
- It requires you to recompile your application every time you want to make a Shiro configuration change.
However, even with these caveats, the direct programmatic manipulation approach could still be valuable in memory-constrained environments, like smart-phone applications. If your application does not run in a memory-constrained environment, you'll find text-based configuration to be easier to use and read.
Most applications instead benefit from text-based configuration that could be modified independently of source code and even make things easier to understand for those not intimately familiar with Shiro's APIs.
To ensure a common-denominator text-based configuration mechanism that can work in all environments with minimal 3rd party dependencies, Shiro supports the INI format to build the SecurityManager object graph and its supporting components. INI is easy to read, easy to configure, and is simple to set-up and suits most applications well.
Here are two examples of how to build a SecurityManager based on INI configuration.
We can create the SecurityManager instance from an INI resource path. Resources can be acquired from the file system, classpath, or URLs when prefixed with file:, classpath:, or url: respectively. This example uses a Factory to ingest a shiro.ini file from the root of the classpath and return the SecurityManager instance:
The INI configuration can be constructed programmatically as well if desired via the org.apache.shiro.config.Ini class. The Ini class functions similarly to the JDK java.util.Properties class, but additionally supports segmentation by section name.
Now that we know how to construct a SecurityManager from INI configuration, let's take a look at exactly how to define a Shiro INI configuration.
INI is basically a text configuration consisting of key/value pairs organized by uniquely-named sections. Keys are unique per section only, not over the entire configuration (unlike the JDK Properties). Each section may be viewed like a single Properties definition however.
Commented lines can start with either with an Octothorpe (# - aka the 'hash', 'pound' or 'number' sign) or a Semi-colon (';')
Here is an example of the sections understood by Shiro:
The [main] section is where you configure the application's SecurityManager instance and any of its dependencies, such as Realms.
Configuring object instances like the SecurityManager or any of its dependencies sounds like a difficult thing to do with INI, where we can only use name/value pairs. But through a little bit of convention and understanding of object graphs, you'll find that you can do quite a lot. Shiro uses these assumptions to enable a simple yet fairly concise configuration mechanism.
We often like to refer to this approach as "poor man's" Dependency Injection, and although not as powerful as full-blown Spring/Guice/JBoss XML files, you'll find it gets quite a lot done without much complexity. Of course those other configuration mechanism are available as well, but they're not required to use Shiro.
Just to whet your appetite, here is an example of a valid [main] configuration. We'll cover it in detail below, but you might find that you understand quite a bit of what is going on already by intuition alone:
Consider the following [main] section snippet:
This line instantiates a new object instance of type com.company.shiro.realm.MyRealm and makes that object available under the myRealm name for further reference and configuration.
If the object instantiated implements the org.apache.shiro.util.Nameable interface, then the the Nameable.setName method will be invoked on the object with the name value (myRealm in this example).
Simple primitive properties can be assigned just by using the equals sign:
these lines of configuration translate into method calls:
Under the covers, Shiro by default uses Apache Commons BeanUtils to do all the heavy lifting when setting these properties. So although INI values are text, BeanUtils knows how to convert the string values to the proper primitive types and then invoke the corresponding JavaBeans setter method.
What if the value you need to set is not a primitive, but another object? Well, you can use a dollar sign ($) to reference a previously-defined instance. For example:
This simply locates the object defined by the name sha256Matcher and then uses BeanUtils to set that object on the myRealm instance (by calling the myRealm.setCredentialsMatcher(sha256Matcher) method).
Using dotted notation on the left side of the INI line's equals sign, you can traverse an object graph to get to the final object/property that you want set. For example, this config line:
Translates (by BeanUtils) into the following logic:
The graph traversal can be as deep as necessary: object.property1.property2....propertyN.value = blah
|BeanUtils Property Support|
Any property assignment operation supported by the BeanUtils.setProperty method will work in Shiro's [main] section, including set/list/map element assignments. See the Apache Commons BeanUtils Website and documentation for more information.
Because raw byte arrays can't be specified natively in a text format, we must use a text encoding of the byte array. The values can be specified either as a Base64 encoded string (the default) or as a Hex encoded string. The default is Base64 because Base64 encoding requires less actual text to represent values - it has a larger encoding alphabet, meaning your tokens are shorter (a bit nicer for text config).
However, if you prefer to use Hex encoding instead, you must prefix the String token with 0x ('zero' 'x'):
Lists, Sets and Maps can be set like any other property - either directly or as a nested property. For sets and lists, just specify a comma-delimited set of values or object references.
For example, some SessionListeners:
For Maps, you specify a comma-delimited list of key-value pairs, where each key-value pair is delimited by a colon ':'
In the above example, the object referenced by $object1 will be in the map under the String key key1, i.e. map.get("key1") returns object1. You can also use other objects as the keys:
The INI format and conventions above are very convenient and easy to understand, but it is not as powerful as other text/XML-based configuration mechanisms. The most important thing to understand when using the above mechanism is that Order Matters!
Each object instantiation and each value assignment is executed in the order they occur in the [main] section. These lines ultimately translate to a JavaBeans getter/setter method invocation, and so those methods are invoked in the same order!
Keep this in mind when writing your configuration.
Any object can be overridden by a new instance defined later in the configuration. So for example, the 2nd myRealm definition would overwrite the first:
This would result in myRealm being a com.company.security.DatabaseRealm instance and the previous instance will never be used (and garbage collected).
You may have noticed in the complete example above that the SecurityManager instance's class isn't defined, and we jumped right in to just setting a nested property:
This is because the securityManager instance is a special one - it is already instantiated for you and ready to go so you don't need to know the specific SecurityManager implementation class to instantiate.
Of course, if you actually want to specify your own implementation, you can, just define your implementation as specified in the "Overriding Instances" section above:
Of course, this is rarely needed - Shiro's SecurityManager implementations are very customizable and can typically be configured with anything necessary. You might want to ask yourself (or the user list) if you really need to do this.
The [users] section allows you to define a static set of user accounts. This is mostly useful in environments with a very small number of user accounts or where user accounts don't need to be created dynamically at runtime. Here's an example:
Just defining non-empty [users] or [roles] sections will automatically trigger the creation of an org.apache.shiro.realm.text.IniRealm instance and make it available in the [main] section under the name iniRealm. You can configure it like any other object as described above.
Each line in the [users] section must conform to the following format:
username = password, roleName1, roleName2, ..., roleNameN
- The value on the left of the equals sign is the username
- The first value on the right of the equals sign is the user's password. A password is required.
- Any comma-delimited values after the password are the names of roles assigned to that user. Role names are optional.
If you don't want the [users] section passwords to be in plain-text, you can encrypt them using your favorite hash algorithm (MD5, Sha1, Sha256, etc) however you like and use the resulting string as the password value. By default, the password string is expected to be Hex encoded, but can be configured to be Base64 encoded instead (see below).
|Easy Secure Passwords|
To save time and use best-practices, you might want to use Shiro's Command Line Hasher, which will hash passwords as well as any other type of resource. It is especially convenient for encrypting INI [users] passwords.
Once you've specified the hashed text password values, you have to tell Shiro that these are encrypted. You do that by configuring the implicitly created iniRealm in the [main] section to use an appropriate CredentialsMatcher implementation corresponding to the hash algorithm you've specified:
You can configure any properties on the CredentialsMatcher like any other object to reflect your hashing strategy, for example, to specify if salting is used or how many hash iterations to execute. See the org.apache.shiro.authc.credential.HashedCredentialsMatcher JavaDoc to better understand hashing strategies and if they might be useful to you.
For example, if your users' password strings were Base64 encoded instead of the default Hex, you could specify:
The [roles] section allows you to associate Permissions with the roles defined in the [users] section. Again, this is useful in environments with a small number of roles or where roles don't need to be created dynamically at runtime. Here's an example:
Each line in the [roles] section must must define a role-to-permission(s) key/value mapping with in the following format:
rolename = permissionDefinition1, permissionDefinition2, ..., permissionDefinitionN
where permissionDefinition is an arbitrary String, but most people will want to use strings that conform
to the org.apache.shiro.authz.permission.WildcardPermission format for ease of use and flexibility. See the Permissions documentation for more information on Permissions and how you can benefit from them.
Note that if an individual permissionDefinition needs to be internally comma-delimited (e.g. printer:5thFloor:print,info), you will need to surround that definition with double quotes (") to avoid parsing errors:
|Roles without Permissions|
If you have roles that don't require permission associations, you don't need to list them in the [roles] section if you don't want to. Just defining the role names in the [users] section is enough to create the role if it does not exist yet.
This section and its options is described in the Web chapter.
While we hope this documentation helps you with the work you're doing with Apache Shiro, the community is improving and expanding the documentation all the time. If you'd like to help the Shiro project, please consider corrected, expanding, or adding documentation where you see a need. Every little bit of help you provide expands the community and in turn improves Shiro.