spaCy/website/setup/jinja_to_js.py

1273 lines
46 KiB
Python
Raw Normal View History

# Forked from: https://github.com/jonbretman/jinja-to-js
2020-08-15 15:50:29 +03:00
# With additional functionality: in/not in, replace, pprint, round, + for lists,
# rendering empty dicts
# This script is mostly used to generate the JavaScript function for the
2020-08-17 15:05:48 +03:00
# training quickstart widget.
import contextlib
import json
import re
import os
from os import path
from io import StringIO
from jinja2 import Environment, FileSystemLoader, nodes
from pathlib import Path
import srsly
import sys
OPERANDS = {
"eq": "===",
"ne": "!==",
"lt": " < ",
"gt": " > ",
"lteq": " <= ",
"gteq": " >= ",
}
DICT_ITER_METHODS = ("iteritems", "items", "values", "keys")
STATE_DEFAULT = 0
STATE_EXECUTING = 1
STATE_INTERPOLATING = 2
LOOP_HELPER_INDEX = "index"
LOOP_HELPER_INDEX_0 = "index0"
LOOP_HELPER_FIRST = "first"
LOOP_HELPER_LAST = "last"
LOOP_HELPER_LENGTH = "length"
LOOP_HELPERS = (
LOOP_HELPER_INDEX,
LOOP_HELPER_INDEX_0,
LOOP_HELPER_FIRST,
LOOP_HELPER_LAST,
LOOP_HELPER_LENGTH,
)
def amd_format(dependencies, template_function):
result = "define(["
result += ",".join('"{0}"'.format(x[0]) for x in dependencies)
result += "], function ("
result += ",".join(x[1] for x in dependencies)
result += ") { return "
result += template_function
result += "; });"
return result
def commonjs_format(dependencies, template_function):
result = "".join('var {0} = require("{1}");'.format(y, x) for x, y in dependencies)
result += "module.exports = {0};".format(template_function)
return result
def es6_format(dependencies, template_function):
result = "".join('import {0} from "{1}";'.format(y, x) for x, y in dependencies)
result += "export default {0}".format(template_function)
return result
JS_MODULE_FORMATS = {
None: lambda dependencies, template_function: template_function,
"amd": amd_format,
"commonjs": commonjs_format,
"es6": es6_format,
}
# This string has to double all the '{' and '}' due to Python's string formatting.
# See - https://docs.python.org/2/library/string.html#formatstrings
TEMPLATE_WRAPPER = """
function {function_name}(ctx) {{
var __result = "";
var __tmp;
var __runtime = jinjaToJS.runtime;
var __filters = jinjaToJS.filters;
var __globals = jinjaToJS.globals;
var context = jinjaToJS.createContext(ctx);
{template_code}
return __result;
}}
"""
class ExtendsException(Exception):
"""
Raised when an {% extends %} is encountered. At this point the parent template is
loaded and all blocks defined in the current template passed to it.
"""
pass
@contextlib.contextmanager
def option(current_kwargs, **kwargs):
"""
Context manager for temporarily setting a keyword argument and
then restoring it to whatever it was before.
"""
tmp_kwargs = dict((key, current_kwargs.get(key)) for key, value in kwargs.items())
current_kwargs.update(kwargs)
yield
current_kwargs.update(tmp_kwargs)
def is_method_call(node, method_name):
"""
Returns True if `node` is a method call for `method_name`. `method_name`
can be either a string or an iterable of strings.
"""
if not isinstance(node, nodes.Call):
return False
if isinstance(node.node, nodes.Getattr):
# e.g. foo.bar()
method = node.node.attr
elif isinstance(node.node, nodes.Name):
# e.g. bar()
method = node.node.name
elif isinstance(node.node, nodes.Getitem):
# e.g. foo["bar"]()
method = node.node.arg.value
else:
return False
if isinstance(method_name, (list, tuple)):
return method in method_name
return method == method_name
def is_loop_helper(node):
"""
Returns True is node is a loop helper e.g. {{ loop.index }} or {{ loop.first }}
"""
return (
hasattr(node, "node")
and isinstance(node.node, nodes.Name)
and node.node.name == "loop"
)
def temp_var_names_generator():
x = 0
while True:
yield "__$%s" % x
x += 1
class JinjaToJS(object):
def __init__(
self,
template_root,
template_name,
js_module_format=None,
runtime_path="jinja-to-js",
include_prefix="",
include_ext="",
child_blocks=None,
dependencies=None,
custom_filters=None,
):
"""
Args:
template_root (str): The path to where templates should be loaded from.
template_name (str): The name of the template to compile (relative to `template_root`).
js_module_format (str, optional): The JavaScript module format to use.
One of ('amd', 'commonjs', 'es6')
runtime_path (str, optional): If `js_module_format` is specified then the JavaScript
runtime will be imported using the appropriate method.
It defaults to assuming it will be imported from
`node_modules` but you can change it using this option.
include_prefix (str, optional): If using the `amd` module format you can use this option
to add a prefix to every include path as AMD imports are
generally relative to the main file, not the module
importing.
include_ext (str, optional): By default any includes will be references without an
extension, as neither AMD, commonJS or ES6 require the
'.js' extension. If you want to use an extension, say
'.template' then set this option to a string including
the leading '.'
child_blocks (dict, optional): Used internally when handling templates that extend
other templates.
dependencies (list of tuple, optional): Used internally when handling templates that
extend other templates.
custom_filters (list of str, optional): List of custom filters which should be allowed.
These may be filters supported by Jinja but not
supported by jinja-to-js. These filters MUST be
registered with the jinja-to-js JS runtime.
"""
self.environment = Environment(
loader=FileSystemLoader(template_root),
autoescape=True,
)
self.output = StringIO()
self.stored_names = set()
self.temp_var_names = temp_var_names_generator()
self.state = STATE_DEFAULT
self.child_blocks = child_blocks or {}
self.dependencies = dependencies or []
self._runtime_function_cache = []
self.js_module_format = js_module_format
self.runtime_path = runtime_path
self.include_prefix = include_prefix
self.include_ext = include_ext
self.template_root = template_root
self.template_name = template_name
self.custom_filters = custom_filters or []
# The name of the JavaScript function that will output this template. By using a named
# function the template can call itself which is required to support recursive includes.
self.js_function_name = "template" + "".join(
x.title()
for x in re.split(r"[^\w]|_", path.splitext(self.template_name)[0])
)
self.context_name = "context"
self._add_dependency(self.runtime_path, "jinjaToJS")
# Jinja2 doesn't accept Windows filepaths
if os.name == "nt":
self.template_name = self.template_name.replace(os.pathsep, "/")
template_string, template_path, _ = self.environment.loader.get_source(
self.environment, self.template_name
)
# It is assumed that this will be the absolute path to the template. It is used to work out
# related paths for inclues.
self.template_path = template_path
if self.js_module_format not in JS_MODULE_FORMATS.keys():
raise ValueError(
"The js_module_format option must be one of: %s"
% JS_MODULE_FORMATS.keys()
)
self.ast = self.environment.parse(template_string)
try:
for node in self.ast.body:
self._process_node(node)
except ExtendsException:
pass
def get_output(self):
"""
Returns the generated JavaScript code.
Returns:
str
"""
# generate the JS function string
template_function = TEMPLATE_WRAPPER.format(
function_name=self.js_function_name, template_code=self.output.getvalue()
).strip()
# get the correct module format template
module_format = JS_MODULE_FORMATS[self.js_module_format]
# generate the module code
return module_format(self.dependencies, template_function)
def _get_depencency_var_name(self, dependency):
"""
Returns the variable name assigned to the given dependency or None if the dependency has
not yet been registered.
Args:
dependency (str): Thet dependency that needs to be imported.
Returns:
str or None
"""
for dep_path, var_name in self.dependencies:
if dep_path == dependency:
return var_name
def _add_dependency(self, dependency, var_name=None):
"""
Adds the given dependency and returns the variable name to use to access it. If `var_name`
is not given then a random one will be created.
Args:
dependency (str):
var_name (str, optional):
Returns:
str
"""
if var_name is None:
var_name = next(self.temp_var_names)
# Don't add duplicate dependencies
if (dependency, var_name) not in self.dependencies:
self.dependencies.append((dependency, var_name))
return var_name
def _process_node(self, node, **kwargs):
node_name = node.__class__.__name__.lower()
handler = getattr(self, "_process_" + node_name, None)
if callable(handler):
handler(node, **kwargs)
else:
2020-08-15 15:50:29 +03:00
raise Exception(f"Unknown node {node} ({node_name})")
def _process_extends(self, node, **kwargs):
"""
Processes an extends block e.g. `{% extends "some/template.jinja" %}`
"""
# find all the blocks in this template
for b in self.ast.find_all(nodes.Block):
# if not already in `child_blocks` then this is the first time a
# block with this name has been encountered.
if b.name not in self.child_blocks:
self.child_blocks[b.name] = b
else:
# otherwise we have seen this block before, so we need to find the last
# super_block and add the block from this template to the end.
block = self.child_blocks.get(b.name)
while hasattr(block, "super_block"):
block = block.super_block
block.super_block = b
# load the parent template
parent_template = JinjaToJS(
template_root=self.template_root,
template_name=node.template.value,
js_module_format=self.js_module_format,
runtime_path=self.runtime_path,
include_prefix=self.include_prefix,
include_ext=self.include_ext,
child_blocks=self.child_blocks,
dependencies=self.dependencies,
)
# add the parent templates output to the current output
self.output.write(parent_template.output.getvalue())
# Raise an exception so we stop parsing this template
raise ExtendsException
def _process_block(self, node, **kwargs):
"""
Processes a block e.g. `{% block my_block %}{% endblock %}`
"""
# check if this node already has a 'super_block' attribute
if not hasattr(node, "super_block"):
# since it doesn't it must be the last block in the inheritance chain
node.super_block = None
# see if there has been a child block defined - if there is this
# will be the first block in the inheritance chain
child_block = self.child_blocks.get(node.name)
if child_block:
# we have child nodes so we need to set `node` as the
# super of the last one in the chain
last_block = child_block
while hasattr(last_block, "super_block"):
last_block = child_block.super_block
# once we have found it, set this node as it's super block
last_block.super_block = node
# this is the node we want to process as it's the first in the inheritance chain
node = child_block
# process the block passing the it's super along, if this block
# calls super() it will be handled by `_process_call`
for n in node.body:
self._process_node(n, super_block=node.super_block, **kwargs)
def _process_output(self, node, **kwargs):
"""
Processes an output node, which will contain things like `Name` and `TemplateData` nodes.
"""
for n in node.nodes:
self._process_node(n, **kwargs)
def _process_templatedata(self, node, **_):
"""
Processes a `TemplateData` node, this is just a bit of as-is text
to be written to the output.
"""
# escape double quotes
value = re.sub('"', r'\\"', node.data)
# escape new lines
value = re.sub("\n", r"\\n", value)
# append value to the result
self.output.write('__result += "' + value + '";')
def _process_name(self, node, **kwargs):
"""
Processes a `Name` node. Some examples of `Name` nodes:
{{ foo }} -> 'foo' is a Name
{% if foo }} -> 'foo' is a Name
"""
with self._interpolation():
with self._python_bool_wrapper(**kwargs):
if node.name not in self.stored_names and node.ctx != "store":
self.output.write(self.context_name)
self.output.write(".")
if node.ctx == "store":
self.stored_names.add(node.name)
self.output.write(node.name)
2020-08-15 15:50:29 +03:00
def _process_dict(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs):
if node.items:
err = f"Can't process non-empty dict in expression: {node}"
raise ValueError(err)
2020-08-15 15:50:29 +03:00
self.output.write("{}")
def _process_getattr(self, node, **kwargs):
"""
Processes a `GetAttr` node. e.g. {{ foo.bar }}
"""
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
if is_loop_helper(node):
self._process_loop_helper(node, **new_kwargs)
else:
self._process_node(node.node, **new_kwargs)
self.output.write(".")
self.output.write(node.attr)
def _process_getitem(self, node, **kwargs):
"""
Processes a `GetItem` node e.g. {{ foo["bar"] }}
"""
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
if isinstance(node.arg, nodes.Slice):
self.output.write(".slice(")
if node.arg.step is not None:
raise Exception(
"The step argument is not supported when slicing."
)
if node.arg.start is None:
self.output.write("0")
else:
self._process_node(node.arg.start, **new_kwargs)
if node.arg.stop is None:
self.output.write(")")
else:
self.output.write(",")
self._process_node(node.arg.stop, **new_kwargs)
self.output.write(")")
else:
self.output.write("[")
self._process_node(node.arg, **new_kwargs)
self.output.write("]")
def _process_for(self, node, **kwargs):
"""
Processes a for loop. e.g.
{% for number in numbers %}
{{ number }}
{% endfor %}
{% for key, value in somemap.items() %}
{{ key }} -> {{ value }}
{% %}
"""
# since a for loop can introduce new names into the context
# we need to remember the ones that existed outside the loop
previous_stored_names = self.stored_names.copy()
with self._execution():
self.output.write("__runtime.each(")
if is_method_call(node.iter, dict.keys.__name__):
self.output.write("Object.keys(")
self._process_node(node.iter, **kwargs)
if is_method_call(node.iter, dict.keys.__name__):
self.output.write(")")
self.output.write(",")
self.output.write("function")
self.output.write("(")
# javascript iterations put the value first, then the key
if isinstance(node.target, nodes.Tuple):
if len(node.target.items) > 2:
raise Exception(
"De-structuring more than 2 items is not supported."
)
for i, item in enumerate(reversed(node.target.items)):
self._process_node(item, **kwargs)
if i < len(node.target.items) - 1:
self.output.write(",")
else:
self._process_node(node.target, **kwargs)
self.output.write(")")
self.output.write("{")
if node.test:
self.output.write("if (!(")
self._process_node(node.test, **kwargs)
self.output.write(")) { return; }")
assigns = (
node.target.items if isinstance(node.target, nodes.Tuple) else [node.target]
)
with self._scoped_variables(assigns, **kwargs):
for n in node.body:
self._process_node(n, **kwargs)
with self._execution():
self.output.write("}")
self.output.write(")")
self.output.write(";")
# restore the stored names
self.stored_names = previous_stored_names
def _process_if(self, node, execute_end=None, **kwargs):
"""
Processes an if block e.g. `{% if foo %} do something {% endif %}`
"""
with self._execution():
self.output.write("if")
self.output.write("(")
with option(kwargs, use_python_bool_wrapper=True):
self._process_node(node.test, **kwargs)
self.output.write(")")
self.output.write("{")
# We accept an `execute_end` function as a keyword argument as this function is
# recursive in the case of something like if-elif-elif-else. In these cases this
# invocation of this function may have to close execution opened by a previous
# invocation of this function.
if execute_end:
execute_end()
# body
for n in node.body:
self._process_node(n, **kwargs)
if not node.else_ and not node.elif_:
# no else - just close the if
with self._execution():
self.output.write("}")
else:
# either an else or an elif
with self._execution() as execute_end:
self.output.write("}")
self.output.write(" else ")
# check for elif
for n in node.elif_:
self._process_node(n, execute_end=execute_end, **kwargs)
if node.elif_ and node.else_:
self.output.write(" else ")
# open up the body
self.output.write("{")
# process the body of the else
for n in node.else_:
self._process_node(n, **kwargs)
# close the body
with self._execution():
self.output.write("}")
def _process_condexpr(self, node, **kwargs):
with self._interpolation():
self.output.write("(")
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.test, **new_kwargs)
self.output.write(" ? ")
self._process_node(node.expr1, **kwargs)
self.output.write(" : ")
self._process_node(node.expr2, **kwargs)
self.output.write(")")
def _process_not(self, node, **kwargs):
self.output.write("!")
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
def _process_or(self, node, **kwargs):
self._process_node(node.left, **kwargs)
self.output.write(" || ")
self._process_node(node.right, **kwargs)
def _process_and(self, node, **kwargs):
self._process_node(node.left, **kwargs)
self.output.write(" && ")
self._process_node(node.right, **kwargs)
def _process_tuple(self, node, **kwargs):
self.output.write("[")
for i, item in enumerate(node.items):
self._process_node(item, **kwargs)
if i < len(node.items) - 1:
self.output.write(",")
self.output.write("]")
def _process_call(self, node, super_block=None, **kwargs):
if is_method_call(node, DICT_ITER_METHODS):
# special case for dict methods
self._process_node(node.node.node, **kwargs)
elif is_method_call(node, "super"):
# special case for the super() method which is available inside blocks
if not super_block:
raise Exception("super() called outside of a block with a parent.")
self._process_node(super_block, **kwargs)
else:
# just a normal function call on a context variable
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
self.output.write("(")
self._process_args(node, **new_kwargs)
self.output.write(")")
# only output the semi-colon if we are not interpolating
if self.state != STATE_INTERPOLATING:
self.output.write("")
def _process_filter(self, node, **kwargs):
method_name = getattr(self, "_process_filter_%s" % node.name, None)
if callable(method_name):
method_name(node, **kwargs)
elif node.name in self.custom_filters:
with self._interpolation(safe=True):
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.%s(" % node.name)
self._process_node(node.node, **new_kwargs)
if getattr(node, "args", None):
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
else:
raise Exception("Unsupported filter: %s" % node.name)
def _process_filter_safe(self, node, **kwargs):
with self._interpolation(safe=True):
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
def _process_filter_capitalize(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.capitalize(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_abs(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("Math.abs(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
2020-08-15 15:50:29 +03:00
def _process_filter_replace(self, node, **kwargs):
# We're getting a quoted string from Python/Jinja as the pattern to
# replace, but to replace all occurrences in JS, we typically need a
# regex, which would be annoying to convert. So we're using split/join
# instead here.
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
self.output.write(".split(")
self._process_node(node.args[0], **new_kwargs)
self.output.write(").join(")
self._process_node(node.args[1], **new_kwargs)
self.output.write(")")
def _process_filter_pprint(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("JSON.stringify(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_attr(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.node, **new_kwargs)
self.output.write("[")
self._process_node(node.args[0], **new_kwargs)
self.output.write("]")
def _process_filter_batch(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.batch(")
self._process_node(node.node, **new_kwargs)
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
def _process_filter_default(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.default(")
self._process_node(node.node, **new_kwargs)
if node.args:
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
def _process_filter_first(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.first(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_int(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.int(")
self._process_node(node.node, **new_kwargs)
if node.args:
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
def _process_filter_round(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("Math.round((")
self._process_node(node.node, **new_kwargs)
2020-08-15 15:50:29 +03:00
self.output.write("+ Number.EPSILON) * 10**")
self._process_node(node.args[0], **new_kwargs)
self.output.write(") / 10**")
self._process_node(node.args[0], **new_kwargs)
def _process_filter_last(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.last(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_length(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.size(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_lower(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("(")
self._process_node(node.node, **new_kwargs)
self.output.write(' + "").toLowerCase()')
def _process_filter_slice(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.slice(")
self._process_node(node.node, **new_kwargs)
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
def _process_filter_title(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.title(")
self._process_node(node.node, **new_kwargs)
self.output.write(")")
def _process_filter_trim(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("(")
self._process_node(node.node, **new_kwargs)
self.output.write(' + "").trim()')
def _process_filter_upper(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("(")
self._process_node(node.node, **new_kwargs)
self.output.write(' + "").toUpperCase()')
def _process_filter_truncate(self, node, **kwargs):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self.output.write("__filters.truncate(")
self._process_node(node.node, **new_kwargs)
self.output.write(",")
self._process_args(node, **new_kwargs)
self.output.write(")")
def _process_assign(self, node, **kwargs):
with self._execution():
self.output.write("var ")
self._process_node(node.target, **kwargs)
self.output.write(" = ")
self._process_node(node.node, **kwargs)
self.output.write(";")
def _process_with(self, node, **kwargs):
# keep a copy of the stored names before the scope
previous_stored_names = self.stored_names.copy()
# assigns in the with tag
# e.g. {% with var = "something %}
assigns_in_tag = [nodes.Assign(t, v) for t, v in zip(node.targets, node.values)]
# assigns in the with body
# e.g. {% set name = 'John' %}
assigns_in_body = [x for x in node.body if isinstance(x, nodes.Assign)]
# remove assigns from the body
node.body = [x for x in node.body if not isinstance(x, nodes.Assign)]
# get a list of all the assigns in this with block
# both on the tag, and within the body of the block
all_assigns = assigns_in_tag + assigns_in_body
with self._execution():
self.output.write("(function () {")
with self._scoped_variables(all_assigns, **kwargs):
for node in node.body:
self._process_node(node, **kwargs)
with self._execution():
self.output.write("})();")
# restore previous stored names
self.stored_names = previous_stored_names
def _process_compare(self, node, **kwargs):
if len(node.ops) > 1:
raise Exception("Multiple operands are not supported.")
operand = node.ops[0]
is_equality = operand.op in ("eq", "ne")
left_hand_is_const = isinstance(node.expr, nodes.Const)
right_hand_is_const = isinstance(operand.expr, nodes.Const)
# If the operand is equality and neither the left or right hand side are constants then we
# will need to use the JavaScript deep equals function. Ideally we want to avoid using this
# as it is quite a big function.
use_is_equal_function = is_equality and not (
left_hand_is_const or right_hand_is_const
)
with option(kwargs, use_python_bool_wrapper=False):
2020-08-13 18:38:30 +03:00
if operand.op == "in" or operand.op == "notin":
# Special case for "in" operator
2020-08-13 18:38:30 +03:00
if operand.op == "notin":
self.output.write("!")
self._process_node(operand.expr, **kwargs)
self.output.write(".includes(")
self._process_node(node.expr, **kwargs)
self.output.write(")")
else:
if use_is_equal_function:
if operand.op == "ne":
self.output.write("!")
self.output.write("__runtime.isEqual(")
self._process_node(node.expr, **kwargs)
if use_is_equal_function:
self.output.write(",")
else:
self.output.write(OPERANDS.get(operand.op))
self._process_node(operand.expr, **kwargs)
if use_is_equal_function:
self.output.write(")")
def _process_operand(self, node, **kwargs):
self.output.write(OPERANDS.get(node.op))
self._process_node(node.expr, **kwargs)
def _process_const(self, node, **_):
with self._interpolation():
self.output.write(json.dumps(node.value))
def _process_nonetype(self, node, **_):
with self._interpolation():
self.output.write("null")
def _process_neg(self, node, **kwargs):
with self._interpolation():
self.output.write("-")
self._process_node(node.node, **kwargs)
def _process_list(self, node, **kwargs):
self.output.write("[")
for i, item in enumerate(node.items):
self._process_node(item, **kwargs)
if i < len(node.items) - 1:
self.output.write(",")
self.output.write("]")
def _process_test(self, node, **kwargs):
with option(kwargs, use_python_bool_wrapper=False):
method_name = getattr(self, "_process_test_%s" % node.name, None)
if callable(method_name):
method_name(node, **kwargs)
else:
raise Exception("Unsupported test: %s" % node.name)
def _process_test_defined(self, node, **kwargs):
self.output.write("(typeof ")
self._process_node(node.node, **kwargs)
self.output.write(' !== "undefined")')
def _process_test_undefined(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(" === undefined")
def _process_test_callable(self, node, **kwargs):
self.output.write("__runtime.type(")
self._process_node(node.node, **kwargs)
self.output.write(') === "Function"')
def _process_test_divisibleby(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(" % ")
self._process_node(node.args[0], **kwargs)
self.output.write(" === 0")
def _process_test_even(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(" % 2 === 0")
def _process_test_odd(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(" % 2 === 1")
def _process_test_none(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(" === null")
def _process_test_upper(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(".toUpperCase() === ")
self._process_node(node.node, **kwargs)
def _process_test_lower(self, node, **kwargs):
self._process_node(node.node, **kwargs)
self.output.write(".toLowerCase() === ")
self._process_node(node.node, **kwargs)
def _process_test_string(self, node, **kwargs):
self.output.write("__runtime.type(")
self._process_node(node.node, **kwargs)
self.output.write(') === "String"')
def _process_test_mapping(self, node, **kwargs):
self.output.write("__runtime.type(")
self._process_node(node.node, **kwargs)
self.output.write(') === "Object"')
def _process_test_number(self, node, **kwargs):
self.output.write("(__runtime.type(")
self._process_node(node.node, **kwargs)
self.output.write(') === "Number" && !isNaN(')
self._process_node(node.node, **kwargs)
self.output.write("))")
def _process_include(self, node, **kwargs):
with self._interpolation(safe=True):
include_path = node.template.value
if include_path == self.template_name:
# template is including itself
include_var_name = self.js_function_name
else:
if self.include_prefix:
include_path = self.include_prefix + node.template.value
elif (
self.js_module_format in ("es6", "commonjs",) and self.template_name
):
_, absolute_include_path, _ = self.environment.loader.get_source(
self.environment, node.template.value
)
include_path = os.path.relpath(
absolute_include_path, os.path.dirname(self.template_path)
)
if not include_path.startswith("."):
include_path = "./" + include_path
# Jinja2 doesn't accept Windows filepaths (but does output them!)
if os.name == "nt":
include_path = include_path.replace(os.pathsep, "/")
include_path = path.splitext(include_path)[0] + self.include_ext
include_var_name = self._get_depencency_var_name(include_path)
if not include_var_name:
include_var_name = self._add_dependency(include_path)
if self.js_module_format is None:
self.output.write('jinjaToJS.include("')
self.output.write(include_path)
self.output.write('");')
else:
self.output.write(include_var_name)
self.output.write("(")
self.output.write(self.context_name)
self.output.write(")")
def _process_add(self, node, **kwargs):
2020-08-15 15:50:29 +03:00
# Handle + operator for lists, which behaves differently in JS. Currently
# only works if we have an explicit list node on either side (in which
# case we assume both are lists).
if isinstance(node.left, nodes.List) or isinstance(node.right, nodes.List):
with self._interpolation():
with self._python_bool_wrapper(**kwargs) as new_kwargs:
self._process_node(node.left, **new_kwargs)
self.output.write(".concat(")
self._process_node(node.right, **new_kwargs)
self.output.write(")")
else:
self._process_math(node, math_operator=" + ", **kwargs)
def _process_sub(self, node, **kwargs):
self._process_math(node, math_operator=" - ", **kwargs)
def _process_div(self, node, **kwargs):
self._process_math(node, math_operator=" / ", **kwargs)
def _process_floordiv(self, node, **kwargs):
self._process_math(node, math_operator=" / ", function="Math.floor", **kwargs)
def _process_mul(self, node, **kwargs):
self._process_math(node, math_operator=" * ", **kwargs)
def _process_mod(self, node, **kwargs):
self._process_math(node, math_operator=" % ", **kwargs)
def _process_math(self, node, math_operator=None, function=None, **kwargs):
"""
Processes a math node e.g. `Div`, `Sub`, `Add`, `Mul` etc...
If `function` is provided the expression is wrapped in a call to that function.
"""
with self._interpolation():
if function:
self.output.write(function)
self.output.write("(")
self._process_node(node.left, **kwargs)
self.output.write(math_operator)
self._process_node(node.right, **kwargs)
if function:
self.output.write(")")
def _process_loop_helper(self, node, **kwargs):
"""
Processes a loop helper e.g. {{ loop.first }} or {{ loop.index }}
"""
if node.attr == LOOP_HELPER_INDEX:
self.output.write("(arguments[1] + 1)")
elif node.attr == LOOP_HELPER_INDEX_0:
self.output.write("arguments[1]")
elif node.attr == LOOP_HELPER_FIRST:
self.output.write("(arguments[1] == 0)")
elif node.attr == LOOP_HELPER_LAST:
self.output.write("(arguments[1] == arguments[2].length - 1)")
elif node.attr == LOOP_HELPER_LENGTH:
self.output.write("arguments[2].length")
def _process_args(self, node, **kwargs):
args = getattr(node, "args", None)
if not args:
return
for i, item in enumerate(args):
self._process_node(item, **kwargs)
if i < len(node.args) - 1:
self.output.write(",")
@contextlib.contextmanager
def _execution(self):
"""
Context manager for executing some JavaScript inside a template.
"""
did_start_executing = False
if self.state == STATE_DEFAULT:
did_start_executing = True
self.state = STATE_EXECUTING
def close():
if did_start_executing and self.state == STATE_EXECUTING:
self.state = STATE_DEFAULT
yield close
close()
@contextlib.contextmanager
def _interpolation(self, safe=False):
did_start_interpolating = False
if self.state == STATE_DEFAULT:
did_start_interpolating = True
self.output.write('__result += "" + ')
if safe is not True:
self.output.write("__runtime.escape")
self.output.write("((__tmp = (")
self.state = STATE_INTERPOLATING
def close():
if did_start_interpolating and self.state == STATE_INTERPOLATING:
self.output.write(')) == null ? "" : __tmp);')
self.state = STATE_DEFAULT
yield close
close()
@contextlib.contextmanager
def _scoped_variables(self, nodes_list, **kwargs):
"""
Context manager for creating scoped variables defined by the nodes in `nodes_list`.
These variables will be added to the context, and when the context manager exits the
context object will be restored to it's previous state.
"""
tmp_vars = []
for node in nodes_list:
is_assign_node = isinstance(node, nodes.Assign)
name = node.target.name if is_assign_node else node.name
# create a temp variable name
tmp_var = next(self.temp_var_names)
# save previous context value
with self._execution():
# save the current value of this name
self.output.write(
"var %s = %s.%s;" % (tmp_var, self.context_name, name)
)
# add new value to context
self.output.write("%s.%s = " % (self.context_name, name))
if is_assign_node:
self._process_node(node.node, **kwargs)
else:
self.output.write(node.name)
self.output.write(";")
tmp_vars.append((tmp_var, name))
yield
# restore context
for tmp_var, name in tmp_vars:
with self._execution():
self.output.write("%s.%s = %s;" % (self.context_name, name, tmp_var))
@contextlib.contextmanager
def _python_bool_wrapper(self, **kwargs):
use_python_bool_wrapper = kwargs.get("use_python_bool_wrapper")
if use_python_bool_wrapper:
self.output.write("__runtime.boolean(")
with option(kwargs, use_python_bool_wrapper=False):
yield kwargs
if use_python_bool_wrapper:
self.output.write(")")
def main(template_path, output=None, data_path=None):
"""Convert a jinja2 template to a JavaScript module.
template_path (Path): Path to .jijna file.
output (Optional[Path]): Path to output .js module (stdout if unset).
data_path (Optional[Path]): Optional JSON or YAML file with additional data
to be included in the JS module as the exported variable DATA.
"""
2020-08-15 15:50:29 +03:00
data = "{}"
if data_path is not None:
if data_path.suffix in (".yml", ".yaml"):
data = srsly.read_yaml(data_path)
else:
data = srsly.read_json(data_path)
data = srsly.json_dumps(data) # dump and load for compactness
template_path = Path(template_path)
tpl_file = template_path.parts[-1]
compiler = JinjaToJS(template_path.parent, tpl_file, js_module_format="es6")
header = f"// This file was auto-generated by {__file__} based on {tpl_file}"
2020-08-15 15:50:29 +03:00
data_str = f"export const DATA = {data}"
result = compiler.get_output()
if output is not None:
2020-09-29 15:49:55 +03:00
with output.open("w", encoding="utf8") as f:
2020-08-15 15:50:29 +03:00
f.write(f"{header}\n{result}\n{data_str}")
print(f"Updated {output.parts[-1]}")
else:
print(result)
if __name__ == "__main__":
args = sys.argv[1:]
if not len(args):
raise ValueError("Need at least one argument: path to .jinja template")
template_path = Path(args[0])
output = Path(args[1]) if len(args) > 1 else None
data_path = Path(args[2]) if len(args) > 2 else None
main(template_path, output, data_path)