overview: | Rationale: this special notation was introduced primarily to allow the dynamic loading of partials. The main advantage that this notation offers is to allow dynamic loading of partials, which is particularly useful in cases where polymorphic data needs to be rendered in different ways. Such cases would otherwise be possible to render only with solutions that are convoluted, inefficient, or both. Example. Let's consider the following data: items: [ { content: 'Hello, World!' }, { url: 'http://example.com/foo.jpg' }, { content: 'Some text' }, { content: 'Some other text' }, { url: 'http://example.com/bar.jpg' }, { url: 'http://example.com/baz.jpg' }, { content: 'Last text here' } ] The goal is to render the different types of items in different ways. The items having a key named `content` should be rendered with the template `text.mustache`: {{!text.mustache}} {{content}} And the items having a key named `url` should be rendered with the template `image.mustache`: {{!image.mustache}} There are already several ways to achieve this goal, here below are illustrated and discussed the most significant solutions to this problem. Using Pre-Processing The idea is to use a secondary templating mechanism to dynamically generate the template that will be rendered. The template that our secondary templating mechanism generates might look like this: {{!template.mustache}} {{items.1.content}} {{items.3.content}} {{items.4.content}} {{items.7.content}} This solutions offers the advantages of having more control over the template and minimizing the template blocks to the essential ones. The drawbacks are the rendering speed and the complexity that the secondary templating mechanism requires. Using Lambdas The idea is to inject functions into the data that will be later called from the template. This way the data will look like this: items: [ { content: 'Hello, World!', html: function() { return '{{>text}}'; } }, { url: 'http://example.com/foo.jpg', html: function() { return '{{>image}}'; } }, { content: 'Some text', html: function() { return '{{>text}}'; } }, { content: 'Some other text', html: function() { return '{{>text}}'; } }, { url: 'http://example.com/bar.jpg', html: function() { return '{{>image}}'; } }, { url: 'http://example.com/baz.jpg', html: function() { return '{{>image}}'; } }, { content: 'Last text here', html: function() { return '{{>text}}'; } } ] And the template will look like this: {{!template.mustache}} {{#items}} {{{html}}} {{/items}} The advantage this solution offers is to have a light main template. The drawback is that the data needs to embed logic and template tags in it. Using If-Else Blocks The idea is to put some logic into the main template so it can select the templates at rendering time: {{!template.mustache}} {{#items}} {{#url}} {{>image}} {{/url}} {{#content}} {{>text}} {{/content}} {{/items}} The main advantage of this solution is that it works without adding any overhead fields to the data. It also documents which external templates are appropriate for expansion in this position. The drawback is that this solution isn't optimal for heterogeneous data sets as the main template grows linearly with the number of polymorphic variants. Using Dynamic Names This is the solution proposed by this spec. The idea is to load partials dynamically. This way the data items have to be tagged with the corresponding partial name: items: [ { content: 'Hello, World!', dynamic: 'text' }, { url: 'http://example.com/foo.jpg', dynamic: 'image' }, { content: 'Some text', dynamic: 'text' }, { content: 'Some other text', dynamic: 'text' }, { url: 'http://example.com/bar.jpg', dynamic: 'image' }, { url: 'http://example.com/baz.jpg', dynamic: 'image' }, { content: 'Last text here', dynamic: 'text' } ] And the template would simple look like this: {{!template.mustache}} {{#items}} {{>*dynamic}} {{/items}} Summary: +----------------+---------------------+-----------------------------------+ | Approach | Pros | Cons | +----------------+---------------------+-----------------------------------+ | Pre-Processing | Essential template, | Secondary templating system | | | more control | needed, slower rendering | | Lambdas | Slim template | Data tagging, logic in data | | If Blocks | No data overhead, | Template linear growth | | | self-documenting | | | Dynamic Names | Slim template | Data tagging | +----------------+---------------------+-----------------------------------+ Dynamic Names are a special notation to dynamically determine a tag's content. Dynamic Names MUST be a non-whitespace character sequence NOT containing the current closing delimiter. A Dynamic Name consists of an asterisk, followed by a dotted name. The dotted name follows the same notation as in an Interpolation tag. This tag's dotted name, which is the Dynamic Name excluding the leading asterisk, references a key in the context whose value will be used in place of the Dynamic Name itself as content of the tag. The dotted name resolution produces the same value as an Interpolation tag and does not affect the context for further processing. Set Delimiter tags MUST NOT affect the resolution of a Dynamic Name. The Dynamic Names MUST be resolved against the context stack local to the tag. Failed resolution of the dynamic name SHOULD result in nothing being rendered. Engines that implement Dynamic Names MUST support their use in Partial tags. In engines that also implement the optional inheritance spec, Dynamic Names inside Parent tags SHOULD be supported as well. Dynamic Names cannot be resolved more than once (Dynamic Names cannot be nested). tests: - name: Basic Behavior - Partial desc: The asterisk operator is used for dynamic partials. data: { dynamic: 'content' } template: '"{{>*dynamic}}"' partials: { content: 'Hello, world!' } expected: '"Hello, world!"' - name: Basic Behavior - Name Resolution desc: | The asterisk is not part of the name that will be resolved in the context. data: { dynamic: 'content', '*dynamic': 'wrong' } template: '"{{>*dynamic}}"' partials: { content: 'Hello, world!', wrong: 'Invisible' } expected: '"Hello, world!"' - name: Context Misses - Partial desc: Failed context lookups should be considered falsey. data: { } template: '"{{>*missing}}"' partials: { missing: 'Hello, world!' } expected: '""' - name: Failed Lookup - Partial desc: The empty string should be used when the named partial is not found. data: { dynamic: 'content' } template: '"{{>*dynamic}}"' partials: { foobar: 'Hello, world!' } expected: '""' - name: Context desc: The dynamic partial should operate within the current context. data: { text: 'Hello, world!', example: 'partial' } template: '"{{>*example}}"' partials: { partial: '*{{text}}*' } expected: '"*Hello, world!*"' - name: Dotted Names desc: The dynamic partial should operate within the current context. data: { text: 'Hello, world!', foo: { bar: { baz: 'partial' } } } template: '"{{>*foo.bar.baz}}"' partials: { partial: '*{{text}}*' } expected: '"*Hello, world!*"' - name: Dotted Names - Operator Precedence desc: The dotted name should be resolved entirely before being dereferenced. data: text: 'Hello, world!' foo: 'test' test: bar: baz: 'partial' template: '"{{>*foo.bar.baz}}"' partials: { partial: '*{{text}}*' } expected: '""' - name: Dotted Names - Failed Lookup desc: The dynamic partial should operate within the current context. data: foo: text: 'Hello, world!' bar: baz: 'partial' template: '"{{>*foo.bar.baz}}"' partials: { partial: '*{{text}}*' } expected: '"**"' - name: Dotted names - Context Stacking desc: Dotted names should not push a new frame on the context stack. data: section1: { value: 'section1' } section2: { dynamic: 'partial', value: 'section2' } template: "{{#section1}}{{>*section2.dynamic}}{{/section1}}" partials: partial: '"{{value}}"' expected: '"section1"' - name: Dotted names - Context Stacking Under Repetition desc: Dotted names should not push a new frame on the context stack. data: value: 'test' section1: [ 1, 2 ] section2: { dynamic: 'partial', value: 'section2' } template: "{{#section1}}{{>*section2.dynamic}}{{/section1}}" partials: partial: "{{value}}" expected: "testtest" - name: Dotted names - Context Stacking Failed Lookup desc: Dotted names should resolve against the proper context stack. data: section1: [ 1, 2 ] section2: { dynamic: 'partial', value: 'section2' } template: "{{#section1}}{{>*section2.dynamic}}{{/section1}}" partials: partial: '"{{value}}"' expected: '""""' - name: Recursion desc: Dynamic partials should properly recurse. data: template: 'node' content: 'X' nodes: [ { content: 'Y', nodes: [] } ] template: '{{>*template}}' partials: { node: '{{content}}<{{#nodes}}{{>*template}}{{/nodes}}>' } expected: 'X>' - name: Dynamic Names - Double Dereferencing desc: Dynamic Names can't be dereferenced more than once. data: { dynamic: 'test', 'test': 'content' } template: '"{{>**dynamic}}"' partials: { content: 'Hello, world!' } expected: '""' - name: Dynamic Names - Composed Dereferencing desc: Dotted Names are resolved entirely before dereferencing begins. data: { foo: 'fizz', bar: 'buzz', fizz: { buzz: { content: null } } } template: '"{{>*foo.*bar}}"' partials: { content: 'Hello, world!' } expected: '""' # Whitespace Sensitivity - name: Surrounding Whitespace desc: | A dynamic partial should not alter surrounding whitespace; any whitespace preceding the tag should be treated as indentation while any whitespace succeding the tag should be left untouched. data: { partial: 'foobar' } template: '| {{>*partial}} |' partials: { foobar: "\t|\t" } expected: "| \t|\t |" - name: Inline Indentation desc: | Whitespace should be left untouched: whitespaces preceding the tag should be treated as indentation. data: { dynamic: 'partial', data: '|' } template: " {{data}} {{>*dynamic}}\n" partials: { partial: ">\n>" } expected: " | >\n>\n" - name: Standalone Line Endings desc: '"\r\n" should be considered a newline for standalone tags.' data: { dynamic: 'partial' } template: "|\r\n{{>*dynamic}}\r\n|" partials: { partial: ">" } expected: "|\r\n>|" - name: Standalone Without Previous Line desc: Standalone tags should not require a newline to precede them. data: { dynamic: 'partial' } template: " {{>*dynamic}}\n>" partials: { partial: ">\n>"} expected: " >\n >>" - name: Standalone Without Newline desc: Standalone tags should not require a newline to follow them. data: { dynamic: 'partial' } template: ">\n {{>*dynamic}}" partials: { partial: ">\n>" } expected: ">\n >\n >" - name: Standalone Indentation desc: Each line of the partial should be indented before rendering. data: { dynamic: 'partial', content: "<\n->" } template: | \ {{>*dynamic}} / partials: partial: | | {{{content}}} | expected: | \ | < -> | / # Whitespace Insensitivity - name: Padding Whitespace desc: Superfluous in-tag whitespace should be ignored. data: { dynamic: 'partial', boolean: true } template: "|{{> * dynamic }}|" partials: { partial: "[]" } expected: '|[]|'