Port of mustache.js to julia
Latest commit cdfd446 Jan 6, 2017 @jverzani committed on GitHub Merge pull request #43 from jverzani/elements
introduce vector index notation. Close #42
Failed to load latest commit information.
doc changed mtrender to render, added io object as first argument of render. Jan 17, 2013
src introduce vector index notation Jan 3, 2017
REQUIRE restore Compat, soln didnt play nicely with others Aug 29, 2016



Mustache Mustache

Linux: Build Status   Windows: Build Status

Mustache is

... a logic-less template syntax. It can be used for HTML,
config files, source code - anything. It works by expanding tags in a
template using values provided in a hash or object.

This package ports over most of the mustache.js implementation for use in Julia. All credit should go there. All bugs are my own.


Following the main documentation for Mustache.js we have a "typical Mustache template" defined by:

using Mustache

tpl = mt"""
Hello {{name}}
You have just won {{value}} dollars!
Well, {{taxed_value}} dollars, after taxes.

The values with braces (mustaches on their side) are looked up in a view, such as a dictionary or module. For example,

d = Dict()
d["name"] = "Chris"
d["value"] = 10000
d["taxed_value"] = 10000 - (10000 * 0.4)
d["in_ca"] = true

render(tpl, d)


Hello Chris
You have just won 10000 dollars!

Well, 6000.0 dollars, after taxes.

The render function pieces things together. Like print, the first argument is for an optional IO instance. In the above example, where one is not provided, the sprint function is employed.

The second argument is a either a string or a mustache template. As seen, templates can be made through the mt non-standard string literal. The advantage of using mt, is the template will be processed at compile time so its reuse will be faster.

The templates use tags comprised of matching mustaches ({}), either two or three, to indicate a value to be substituted for.

The third argument is for a view to provide values to substitute into the template. The above example used a dictionary. A Module may also be used, such as Main:

name, value, taxed_value, in_ca = "Christine", 10000, 10000 - (10000 * 0.4), false
render(tpl, Main) |> print

Which yields:

Hello Christine
You have just won 10000 dollars!

Further, keyword arguments can be used when the variables in the templates are symbols:

goes_together = mt"{{{:x}}} and {{{:y}}}."
render(goes_together, x="Salt", y="pepper")
render(goes_together, x="Bread", y="butter")

As well, one can use Composite Kinds. This may make writing show methods easier:

using Distributions
tpl = "Beta distribution with alpha={{α}}, beta={{β}}"
render(tpl, Beta(1, 2))


"Beta distribution with alpha=1.0, beta=2.0"


Tags representing variables have the form {{varname}}, {{:symbol}}, or their triple-braced versions {{{varname}}} or {{{:symbol}}}. The triple brace prevents HTML substitution for entities such as <. The following are escaped when only double braces are used: "&", "<", ">", "'", "\", and "/".


In the main example, the template included:

Well, {{taxed_value}} dollars, after taxes.

Tags beginning with #varname and closed with /varname create sections. The part between them is used only if the variable is defined. Related, if the tag begins with ^varname and ends with /varname the text between these tags is included only if the variable is not defined.


If the section variable, {{#VARNAME}}, binds to an iterable collection, then the text in the section is repeated for each item in the collection with the view used for the context of the template given by the item.

This is useful for collections of named objects, such as DataFrames (where the collection is comprised of rows) or arrays of dictionaries. The special variable {{.}} can be used to iterate over non-named collections.

For data frames the variable names are specified as symbols or strings. Here is a template for making a web page:

tpl = mt"""

This can be used to generate a web page for whos-like values:

_names = Array(String, 0)
_summaries = Array(String, 0)
m = Main
for s in sort(map(string, names(m)))
    v = symbol(s)
    if isdefined(m,v)
        push!(_names, s)
        push!(_summaries, summary(eval(m,v)))

using DataFrames
d = DataFrame(names=_names, summs=_summaries)

out = render(tpl, TITLE="A quick table", D=d)

This can be compared to using an array of Dicts, convenient if you have data by the row:

A = [{"a" => "eh", "b" => "bee"},
     {"a" => "ah", "b" => "buh"}]
tpl = mt"{{#:A}}Pronounce a as {{a}} and b as {{b}}. {{/:A}}"
render(tpl, A=A) |> print


Pronounce a as eh and b as bee. Pronounce a as ah and b as buh.

The same approach can be made to make a LaTeX table from a data frame:

function df_to_table(df, label="label", caption="caption")
    fmt = repeat("c", length(df))
    row = join(["{{$x}}" for x in map(string, names(df))], " & ")

{{#:DF}}    $row\\\\
{{/:DF}}  \\end{tabular}

render(tpl, DF=df)

(A string is used -- and not a mt macro above -- so that string interpolation can happen.)

Iterating over vectors

Iterating over an unnamed vector uses {{.}} to refer to the item:

tpl = "{{#:vec}}{{.}} {{/:vec}}"
render(tpl, vec = ["A1", "B2", "C3"])  # "A1 B2 C3 "

Not the extra space after C3. There is experimental support for indexing with the iteration of a vector that allows on to work around this. The syntax .[ind] refers to the value vec[ind].

To print commas one can use this pattern:

tpl = "{{#:vec}}{{.}}{{^.[end]}}, {{/.[end]}}{{/:vec}}"
render(tpl, vec = ["A1", "B2", "C3"])  # "A1, B2, C3"

To put the first value in bold, but no others, say:

tpl = """
render(tpl, vec = ["A1", "B2", "C3"])  # basically "<bold>A1</bold>B2 C3"

This was inspired by this question, but the syntax chosen was more Julian. This syntax -- as implemented for now -- does not allow for iteration. That is constructs like {{#.[1]}} don't introduce iteration, but only offer a conditional check.


Partials are used to include partial templates into a template.

Partials begin with a greater than sign, like {{> box.tpl }}. In this example, the file box.tpl is opened and inserted into the template, then populated. A full path may be specified.

They also inherit the calling context.

In this way you may want to think of partials as includes, imports, template expansion, nested templates, or subtemplates, even though those aren't literally the case here.

The partial specified by {{< box.tpl }} is not parsed, rather included as is into the file. This can be much faster.


Julia provides some alternatives to this package which are better suited for many jobs:

  • For simple substitution inside a string there is string interpolation.

  • For piecing together pieces of text either the string function or string concatenation (the * operator) are useful.

  • For formatting numbers and text, the Formatting.jl package is available.

Differences from Mustache.js

This project deviates from that of Mustache.js in a few significant ways:

  • The tags are only demarcated with mustaches, this is not customizable
  • Julian structures are used, not JavaScript objects. As illustrated, one can use Dicts, Modules, DataFrames