A Java implementation of the Liquid templating engine backed up by an ANTLR grammar.
Add the dependency:
dependencies {
compile 'nl.big-o:liqp:0.8.3.3'
}
Add the following dependency:
<dependency>
  <groupId>nl.big-o</groupId>
  <artifactId>liqp</artifactId>
  <version>0.8.3.3</version>
</dependency>
Or clone this repository and run: mvn install
which will create a JAR of Liqp
in your local Maven repository, as well as in the project's target/
folder.
This library can be used in two different ways:
- to construct a parse tree of some Liquid input
- to render Liquid input source (either files, or input strings)
To create a parse tree from input source, do the following:
String input =
"<ul id=\"products\"> \n" +
" {% for product in products %} \n" +
" <li> \n" +
" <h2>{{ product.name }}</h2> \n" +
" Only {{ product.price | price }} \n" +
" \n" +
" {{ product.description | prettyprint | paragraph }} \n" +
" </li> \n" +
" {% endfor %} \n" +
"</ul> \n";
Template template = Template.parse(input);
ParseTree root = template.getParseTree();
As you can see, the getParseTree()
method returns an instance of a
ParseTree
denoting the root
node of the input source. To see how the parse tree is built, you can use Template#toStringAST()
to print
an ASCII representation of the tree.
If you're not familiar with Liquid, have a look at their website: http://liquidmarkup.org.
In Ruby, you'd render a template like this:
@template = Liquid::Template.parse("hi {{name}}") # Parses and compiles the template
@template.render( 'name' => 'tobi' ) # Renders the output => "hi tobi"
With Liqp, the equivalent looks like this:
Template template = Template.parse("hi {{name}}");
String rendered = template.render("name", "tobi");
System.out.println(rendered);
/*
hi tobi
*/
The template variables provided as parameters to render(...)
can be:
- a varargs where
the 0th, 2nd, 4th, ... indexes must be
String
literals denoting the keys. The values can be anyObject
. - a
Map<String, Object>
- a JSON string
- any POJO that is marked with special marker interface
liqp.parser.Inspectable
. In this case the object is converted tojava.util.Map
using jackson's mapper, and so all recipes for configuring jackson conversation will work here. - any object that extend special interface
liqp.parser.LiquidSupport
and it is designed for lazy field values computing. It's methodLiquidSupport#toLiquid()
is called only if/when the object is going to be rendered. SinceLiquidSupport
extendsInspectable
simply use same variant of therender(...)
method.
The following examples are equivalent to the previous Liqp example:
Template template = Template.parse("hi {{name}}");
Map<String, Object> map = new HashMap<String, Object>();
map.put("name", "tobi");
String rendered = template.render(map);
System.out.println(rendered);
/*
hi tobi
*/
Template template = Template.parse("hi {{name}}");
String rendered = template.render("{\"name\" : \"tobi\"}");
System.out.println(rendered);
/*
hi tobi
*/
class MyParams implements Inspectable {
public String name = "tobi";
};
Template template = Template.parse("hi {{name}}");
String rendered =template.render(new MyParams());
System.out.println(rendered);
/*
hi tobi
*/
class MyLazy implements LiquidSupport {
@Override
public Map<String, Object> toLiquid() {
return Collections.singletonMap("name", "tobi");
}
};
Template template = Template.parse("hi {{name}}");
String rendered = template.render(new MyLazy());
System.out.println(rendered);
/*
hi tobi
*/
Strict variables means that value for every key must be provided, otherwise an exception occurs.
Template template = Template.parse("hi {{name}}")
.withRenderSettings(new RenderSettings.Builder().withStrictVariables(true).build());
String rendered = template.render(); // no value for "name"
// exception is thrown
There exists two rendering modes: lazy and eager.
- In
lazy
mode the template parameters are evaluating on demand and specific properties are read from there only if they are needed. Each filter/tag trying to do its work with its own parameter object, that can be literally anything. - In
eager
the entire parameter object is converted into plain data tree structure that are made only from maps and lists, so tags/filters do know how to work with these kinds of objects. Special case - temporal objects, they are consumed as is.
By default, the lazy
one is used. This should do the work in most cases.
Switching mode is possible via providing special RenderSettings
.
Example usage of lazy
mode:
RenderSettings renderSettings = new RenderSettings.Builder()
.withEvaluateMode(RenderSettings.EvaluateMode.EAGER)
.build();
Map<String, Object> in = Collections.singletonMap("a", new Object() {
public String val = "tobi";
});
String res = Template.parse("hi {{a.val}}")
.withRenderSettings(renderSettings)
.render(in);
System.out.println(res);
/*
hi tobi
*/
Let's say you want to create a custom filter, called b
, that changes a string like
*text*
to <strong>text</strong>
.
You can do that as follows:
// first register your custom filter
Filter.registerFilter(new Filter("b"){
@Override
public Object apply(Object value, TemplateContext context, Object... params) {
// create a string from the value
String text = super.asString(value, context);
// replace and return *...* with <strong>...</strong>
return text.replaceAll("\\*(\\w(.*?\\w)?)\\*", "<strong>$1</strong>");
}
});
// use your filter
Template template = Template.parse("{{ wiki | b }}");
String rendered = template.render("{\"wiki\" : \"Some *bold* text *in here*.\"}");
System.out.println(rendered);
/*
Some <strong>bold</strong> text <strong>in here</strong>.
*/
And to use an optional parameter in your filter, do something like this:
// first register your custom filter
Filter.registerFilter(new Filter("repeat"){
@Override
public Object apply(Object value, TemplateContext context, Object... params) {
// get the text of the value
String text = super.asString(value, context);
// check if an optional parameter is provided
int times = params.length == 0 ? 1 : super.asNumber(params[0]).intValue();
StringBuilder builder = new StringBuilder();
while(times-- > 0) {
builder.append(text);
}
return builder.toString();
}
});
// use your filter
Template template = Template.parse("{{ 'a' | repeat }}\n{{ 'b' | repeat:5 }}");
String rendered = template.render();
System.out.println(rendered);
/*
a
bbbbb
*/
You can use an array (or list) as well, and can also return a numerical value:
Filter.registerFilter(new Filter("sum"){
@Override
public Object apply(Object value, TemplateContext context, Object... params) {
Object[] numbers = super.asArray(value, context);
double sum = 0;
for(Object obj : numbers) {
sum += super.asNumber(obj).doubleValue();
}
return sum;
}
});
Template template = Template.parse("{{ numbers | sum }}");
String rendered = template.render("{\"numbers\" : [1, 2, 3, 4, 5]}");
System.out.println(rendered);
/*
15.0
*/
In short, override one of the apply()
methods of the Filter
class to create your own custom filter behaviour.
Both blocks and tags are kinds of insertions. Literally: Block extends Insertion
and Tag extends Insertion
. Class Insertion
used as basic abstraction for parser.
Let's define the difference between tags and blocks.
Tag is simple insertion in the template that will be processed and the result of it will be replaced in output, if any. Example is include
tag:
See these data: {% include data.liquid %}
Another example is assign
tag:
{% assign name='Joe' %}
Hello {{name}}!
It has no input but still is an insertion.
Block is a kind of insertions that wraps some text and/or other blocks or tags and perform some operations on the given input. Blocks have opening and closing tags. Example of block is if
:
{% if user %} Hello {{ user.name }} {% endif %}"
where {% if user %}
is opening tag and {% endif %}
is closing one. The user
in this sample is just a parameter for given block.
Let's say you would like to create a block that makes it easy to loop for a fixed amount of times, executing a block of Liquid code.
Here's a way to create, and use, such a custom loop
block:
Block.registerBlock(new Block("loop"){
@Override
public Object render(TemplateContext context, LNode... nodes) {
int n = super.asNumber(nodes[0].render(context)).intValue();
LNode block = nodes[1];
StringBuilder builder = new StringBuilder();
while(n-- > 0) {
builder.append(super.asString(block.render(context), context));
}
return builder.toString();
}
});
Template template = Template.parse("{% loop 5 %}looping!\n{% endloop %}");
String rendered = template.render();
System.out.println(rendered);
/*
looping!
looping!
looping!
looping!
looping!
*/
Note: Block.registerBlock
is just a fluent shortcut for Insertion.registerInsertion
.
For registering custom Tags there exists equivalent Tags.registerTag
function (that also wraps Insertion.registerInsertion
).
Note that all of:
Insertion.registerInsertion(Insertion)
Block.registerBlock(Block)
Tag.registerTag(Tag)
Filer.registerFilter(Filter)
will add tags, blocks and filters per JVM instance. If you want templates to use specific filters, create your Template
instance as follows:
Template.parse(source, new ParseSettings.Builder()
.with(tag)
.build());
Template.parse(source, new ParseSettings.Builder()
.with(block)
.build());
Template.parse(source, new ParseSettings.Builder()
.with(filter)
.build());
// Or combine them:
Template.parse(source, new ParseSettings.Builder()
.with(filter)
.with(tag)
.build());
For example, using the sum
filter for just 1 template, would look like this:
Template template = Template.parse("{{ numbers | sum }}", new ParseSettings.Builder()
.with(new Filter("sum"){
@Override
public Object apply(Object value, TemplateContext context, Object... params) {
Object[] numbers = super.asArray(value, context);
double sum = 0;
for(Object obj : numbers) {
sum += super.asNumber(obj).doubleValue();
}
return sum;
}
}).build());
String rendered = template.render("{\"numbers\" : [1, 2, 3, 4, 5]}");
System.out.println(rendered);
/*
15.0
*/
Also note that ParseSettings
object have unmodifiable fields and so can be shared across templates without side effects.
If you're evaluating templates from untrusted sources, there are a couple of ways you can guard against unwanted input.
For example, if you'd like the input template to be no larger than 125 characters, the templating engine should not perform more than 15 iterations in total, the generated string should not exceed 300 characters and the total rendering (and parsing!) time should not exceed 100 milliseconds, you could do something like this:
ProtectionSettings protectionSettings = new ProtectionSettings.Builder()
.withMaxSizeRenderedString(300)
.withMaxIterations(15)
.withMaxRenderTimeMillis(100L)
.withMaxTemplateSizeBytes(125)
.build();
String rendered = Template.parse("{% for i in (1..10) %}{{ text }}{% endfor %}")
.withProtectionSettings(protectionSettings)
.render("{\"text\": \"abcdefghijklmnopqrstuvwxyz\"}");
System.out.println(rendered);
Note that not providing a ProtectionSettings
, is the same as not having any guards in
place (or better, very large limits).
Use Maven 3.5.0 and run build with
mvn clean install
Release process into the Central Repository is performed with
mvn release:clean release:prepare release:perform -P ossrh-release
Make sure having in ~/.m2/settings.xml
this config(with your values):
<settings>
<servers>
<server>
<id>ossrh</id>
<username>MY_OSSRH_USERNAME</username>
<password>MY_OSSRH_PASSWORD</password>
</server>
</servers>
<profiles>
<profile>
<id>ossrh-release</id>
<activation>
<activeByDefault>false</activeByDefault>
</activation>
<properties>
<gpg.executable>gpg2</gpg.executable>
<gpg.passphrase>GPG_PRIVATE_KEY_PASSWORD</gpg.passphrase>
</properties>
</profile>
</profiles>
</settings>
After executing this go to https://oss.sonatype.org/index.html#stagingRepositories, ensure all is OK, after you can "Close" the staging for promoting to the realease and after do "Release".