blob: df37947df84d5d7820730f0f93f4fd1eca13f62e [file] [log] [blame]
#![allow(deprecated)]
/// The compiler module houses the code which parses and compiles templates. TinyTemplate implements
/// a simple bytecode interpreter (see the [instruction] module for more details) to render templates.
/// The [`TemplateCompiler`](struct.TemplateCompiler.html) struct is responsible for parsing the
/// template strings and generating the appropriate bytecode instructions.
use error::Error::*;
use error::{get_offset, Error, Result};
use instruction::{Instruction, Path, PathStep};
/// The end point of a branch or goto instruction is not known.
const UNKNOWN: usize = ::std::usize::MAX;
/// The compiler keeps a stack of the open blocks so that it can ensure that blocks are closed in
/// the right order. The Block type is a simple enumeration of the kinds of blocks that could be
/// open. It may contain the instruction index corresponding to the start of the block.
enum Block {
Branch(usize),
For(usize),
With,
}
/// List of the known @-keywords so that we can error if the user spells them wrong.
static KNOWN_KEYWORDS: [&str; 4] = ["@index", "@first", "@last", "@root"];
/// The TemplateCompiler struct is responsible for parsing a template string and generating bytecode
/// instructions based on it. The parser is a simple hand-written pattern-matching parser with no
/// recursion, which makes it relatively easy to read.
pub(crate) struct TemplateCompiler<'template> {
original_text: &'template str,
remaining_text: &'template str,
instructions: Vec<Instruction<'template>>,
block_stack: Vec<(&'template str, Block)>,
/// When we see a `{foo -}` or similar, we need to remember to left-trim the next text block we
/// encounter.
trim_next: bool,
}
impl<'template> TemplateCompiler<'template> {
/// Create a new template compiler to parse and compile the given template.
pub fn new(text: &'template str) -> TemplateCompiler<'template> {
TemplateCompiler {
original_text: text,
remaining_text: text,
instructions: vec![],
block_stack: vec![],
trim_next: false,
}
}
/// Consume the template compiler to parse the template and return the generated bytecode.
pub fn compile(mut self) -> Result<Vec<Instruction<'template>>> {
while !self.remaining_text.is_empty() {
// Comment, denoted by {# comment text #}
if self.remaining_text.starts_with("{#") {
self.trim_next = false;
let tag = self.consume_tag("#}")?;
let comment = tag[2..(tag.len() - 2)].trim();
if comment.starts_with('-') {
self.trim_last_whitespace();
}
if comment.ends_with('-') {
self.trim_next_whitespace();
}
// Block tag. Block tags are wrapped in {{ }} and always have one word at the start
// to identify which kind of tag it is. Depending on the tag type there may be more.
} else if self.remaining_text.starts_with("{{") {
self.trim_next = false;
let (discriminant, rest) = self.consume_block()?;
match discriminant {
"if" => {
let (path, negated) = if rest.starts_with("not") {
(self.parse_path(&rest[4..])?, true)
} else {
(self.parse_path(rest)?, false)
};
self.block_stack
.push((discriminant, Block::Branch(self.instructions.len())));
self.instructions
.push(Instruction::Branch(path, !negated, UNKNOWN));
}
"else" => {
self.expect_empty(rest)?;
let num_instructions = self.instructions.len() + 1;
self.close_branch(num_instructions, discriminant)?;
self.block_stack
.push((discriminant, Block::Branch(self.instructions.len())));
self.instructions.push(Instruction::Goto(UNKNOWN))
}
"endif" => {
self.expect_empty(rest)?;
let num_instructions = self.instructions.len();
self.close_branch(num_instructions, discriminant)?;
}
"with" => {
let (path, name) = self.parse_with(rest)?;
let instruction = Instruction::PushNamedContext(path, name);
self.instructions.push(instruction);
self.block_stack.push((discriminant, Block::With));
}
"endwith" => {
self.expect_empty(rest)?;
if let Some((_, Block::With)) = self.block_stack.pop() {
self.instructions.push(Instruction::PopContext)
} else {
return Err(self.parse_error(
discriminant,
"Found a closing endwith that doesn't match with a preceeding with.".to_string()
));
}
}
"for" => {
let (path, name) = self.parse_for(rest)?;
self.instructions
.push(Instruction::PushIterationContext(path, name));
self.block_stack
.push((discriminant, Block::For(self.instructions.len())));
self.instructions.push(Instruction::Iterate(UNKNOWN));
}
"endfor" => {
self.expect_empty(rest)?;
let num_instructions = self.instructions.len() + 1;
let goto_target = self.close_for(num_instructions, discriminant)?;
self.instructions.push(Instruction::Goto(goto_target));
self.instructions.push(Instruction::PopContext);
}
"call" => {
let (name, path) = self.parse_call(rest)?;
self.instructions.push(Instruction::Call(name, path));
}
_ => {
return Err(self.parse_error(
discriminant,
format!("Unknown block type '{}'", discriminant),
));
}
}
// Values, of the form { dotted.path.to.value.in.context }
// Note that it is not (currently) possible to escape curly braces in the templates to
// prevent them from being interpreted as values.
} else if self.remaining_text.starts_with('{') {
self.trim_next = false;
let (path, name) = self.consume_value()?;
let instruction = match name {
Some(name) => Instruction::FormattedValue(path, name),
None => Instruction::Value(path),
};
self.instructions.push(instruction);
// All other text - just consume characters until we see a {
} else {
let mut escaped = false;
loop {
let mut text = self.consume_text(escaped);
if self.trim_next {
text = text.trim_left();
self.trim_next = false;
}
escaped = text.ends_with('\\');
if escaped {
text = &text[0..(text.len() - 1)];
}
self.instructions.push(Instruction::Literal(text));
if !escaped {
break;
}
}
}
}
if let Some((text, _)) = self.block_stack.pop() {
return Err(self.parse_error(
text,
"Expected block-closing tag, but reached the end of input.".to_string(),
));
}
Ok(self.instructions)
}
/// Splits a string into a list of named segments which can later be used to look up values in the
/// context.
fn parse_path(&self, text: &'template str) -> Result<Path<'template>> {
if !text.starts_with('@') {
Ok(text
.split('.')
.map(|s| match s.parse::<usize>() {
Ok(n) => PathStep::Index(s, n),
Err(_) => PathStep::Name(s),
})
.collect::<Vec<_>>())
} else if KNOWN_KEYWORDS.iter().any(|k| *k == text) {
Ok(vec![PathStep::Name(text)])
} else {
Err(self.parse_error(text, format!("Invalid keyword name '{}'", text)))
}
}
/// Finds the line number and column where an error occurred. Location is the substring of
/// self.original_text where the error was found, and msg is the error message.
fn parse_error(&self, location: &str, msg: String) -> Error {
let (line, column) = get_offset(self.original_text, location);
ParseError { msg, line, column }
}
/// Tags which should have no text after the discriminant use this to raise an error if
/// text is found.
fn expect_empty(&self, text: &str) -> Result<()> {
if text.is_empty() {
Ok(())
} else {
Err(self.parse_error(text, format!("Unexpected text '{}'", text)))
}
}
/// Close the branch that is on top of the block stack by setting its target instruction
/// and popping it from the stack. Returns an error if the top of the block stack is not a
/// branch.
fn close_branch(&mut self, new_target: usize, discriminant: &str) -> Result<()> {
let branch_block = self.block_stack.pop();
if let Some((_, Block::Branch(index))) = branch_block {
match &mut self.instructions[index] {
Instruction::Branch(_, _, target) => {
*target = new_target;
Ok(())
}
Instruction::Goto(target) => {
*target = new_target;
Ok(())
}
_ => panic!(),
}
} else {
Err(self.parse_error(
discriminant,
"Found a closing endif or else which doesn't match with a preceding if."
.to_string(),
))
}
}
/// Close the for loop that is on top of the block stack by setting its target instruction and
/// popping it from the stack. Returns an error if the top of the stack is not a for loop.
/// Returns the index of the loop's Iterate instruction for further processing.
fn close_for(&mut self, new_target: usize, discriminant: &str) -> Result<usize> {
let branch_block = self.block_stack.pop();
if let Some((_, Block::For(index))) = branch_block {
match &mut self.instructions[index] {
Instruction::Iterate(target) => {
*target = new_target;
Ok(index)
}
_ => panic!(),
}
} else {
Err(self.parse_error(
discriminant,
"Found a closing endfor which doesn't match with a preceding for.".to_string(),
))
}
}
/// Advance the cursor to the next { and return the consumed text. If `escaped` is true, skips
/// a { at the start of the text.
fn consume_text(&mut self, escaped: bool) -> &'template str {
let search_substr = if escaped {
&self.remaining_text[1..]
} else {
self.remaining_text
};
let mut position = search_substr
.find('{')
.unwrap_or_else(|| search_substr.len());
if escaped {
position += 1;
}
let (text, remaining) = self.remaining_text.split_at(position);
self.remaining_text = remaining;
text
}
/// Advance the cursor to the end of the value tag and return the value's path and optional
/// formatter name.
fn consume_value(&mut self) -> Result<(Path<'template>, Option<&'template str>)> {
let tag = self.consume_tag("}")?;
let mut tag = tag[1..(tag.len() - 1)].trim();
if tag.starts_with('-') {
tag = tag[1..].trim();
self.trim_last_whitespace();
}
if tag.ends_with('-') {
tag = tag[0..tag.len() - 1].trim();
self.trim_next_whitespace();
}
if let Some(index) = tag.find('|') {
let (path_str, name_str) = tag.split_at(index);
let name = name_str[1..].trim();
let path = self.parse_path(path_str.trim())?;
Ok((path, Some(name)))
} else {
Ok((self.parse_path(tag)?, None))
}
}
/// Right-trim whitespace from the last text block we parsed.
fn trim_last_whitespace(&mut self) {
if let Some(Instruction::Literal(text)) = self.instructions.last_mut() {
*text = text.trim_right();
}
}
/// Make a note to left-trim whitespace from the next text block we parse.
fn trim_next_whitespace(&mut self) {
self.trim_next = true;
}
/// Advance the cursor to the end of the current block tag and return the discriminant substring
/// and the rest of the text in the tag. Also handles trimming whitespace where needed.
fn consume_block(&mut self) -> Result<(&'template str, &'template str)> {
let tag = self.consume_tag("}}")?;
let mut block = tag[2..(tag.len() - 2)].trim();
if block.starts_with('-') {
block = block[1..].trim();
self.trim_last_whitespace();
}
if block.ends_with('-') {
block = block[0..block.len() - 1].trim();
self.trim_next_whitespace();
}
let discriminant = block.split_whitespace().next().unwrap_or(block);
let rest = block[discriminant.len()..].trim();
Ok((discriminant, rest))
}
/// Advance the cursor to after the given expected_close string and return the text in between
/// (including the expected_close characters), or return an error message if we reach the end
/// of a line of text without finding it.
fn consume_tag(&mut self, expected_close: &str) -> Result<&'template str> {
if let Some(line) = self.remaining_text.lines().next() {
if let Some(pos) = line.find(expected_close) {
let (tag, remaining) = self.remaining_text.split_at(pos + expected_close.len());
self.remaining_text = remaining;
Ok(tag)
} else {
Err(self.parse_error(
line,
format!(
"Expected a closing '{}' but found end-of-line instead.",
expected_close
),
))
}
} else {
Err(self.parse_error(
self.remaining_text,
format!(
"Expected a closing '{}' but found end-of-text instead.",
expected_close
),
))
}
}
/// Parse a with tag to separate the value path from the (optional) name.
fn parse_with(&self, with_text: &'template str) -> Result<(Path<'template>, &'template str)> {
if let Some(index) = with_text.find(" as ") {
let (path_str, name_str) = with_text.split_at(index);
let path = self.parse_path(path_str.trim())?;
let name = name_str[" as ".len()..].trim();
Ok((path, name))
} else {
Err(self.parse_error(
with_text,
format!(
"Expected 'as <path>' in with block, but found \"{}\" instead",
with_text
),
))
}
}
/// Parse a for tag to separate the value path from the name.
fn parse_for(&self, for_text: &'template str) -> Result<(Path<'template>, &'template str)> {
if let Some(index) = for_text.find(" in ") {
let (name_str, path_str) = for_text.split_at(index);
let name = name_str.trim();
let path = self.parse_path(path_str[" in ".len()..].trim())?;
Ok((path, name))
} else {
Err(self.parse_error(
for_text,
format!("Unable to parse for block text '{}'", for_text),
))
}
}
/// Parse a call tag to separate the template name and context value.
fn parse_call(&self, call_text: &'template str) -> Result<(&'template str, Path<'template>)> {
if let Some(index) = call_text.find(" with ") {
let (name_str, path_str) = call_text.split_at(index);
let name = name_str.trim();
let path = self.parse_path(path_str[" with ".len()..].trim())?;
Ok((name, path))
} else {
Err(self.parse_error(
call_text,
format!("Unable to parse call block text '{}'", call_text),
))
}
}
}
#[cfg(test)]
mod test {
use super::*;
use instruction::Instruction::*;
fn compile(text: &'static str) -> Result<Vec<Instruction<'static>>> {
TemplateCompiler::new(text).compile()
}
#[test]
fn test_compile_literal() {
let text = "Test String";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(&Literal(text), &instructions[0]);
}
#[test]
fn test_compile_value() {
let text = "{ foobar }";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(&Value(vec![PathStep::Name("foobar")]), &instructions[0]);
}
#[test]
fn test_compile_value_with_formatter() {
let text = "{ foobar | my_formatter }";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(
&FormattedValue(vec![PathStep::Name("foobar")], "my_formatter"),
&instructions[0]
);
}
#[test]
fn test_dotted_path() {
let text = "{ foo.bar }";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(
&Value(vec![PathStep::Name("foo"), PathStep::Name("bar")]),
&instructions[0]
);
}
#[test]
fn test_indexed_path() {
let text = "{ foo.0.bar }";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(
&Value(vec![
PathStep::Name("foo"),
PathStep::Index("0", 0),
PathStep::Name("bar")
]),
&instructions[0]
);
}
#[test]
fn test_mixture() {
let text = "Hello { name }, how are you?";
let instructions = compile(text).unwrap();
assert_eq!(3, instructions.len());
assert_eq!(&Literal("Hello "), &instructions[0]);
assert_eq!(&Value(vec![PathStep::Name("name")]), &instructions[1]);
assert_eq!(&Literal(", how are you?"), &instructions[2]);
}
#[test]
fn test_if_endif() {
let text = "{{ if foo }}Hello!{{ endif }}";
let instructions = compile(text).unwrap();
assert_eq!(2, instructions.len());
assert_eq!(
&Branch(vec![PathStep::Name("foo")], true, 2),
&instructions[0]
);
assert_eq!(&Literal("Hello!"), &instructions[1]);
}
#[test]
fn test_if_not_endif() {
let text = "{{ if not foo }}Hello!{{ endif }}";
let instructions = compile(text).unwrap();
assert_eq!(2, instructions.len());
assert_eq!(
&Branch(vec![PathStep::Name("foo")], false, 2),
&instructions[0]
);
assert_eq!(&Literal("Hello!"), &instructions[1]);
}
#[test]
fn test_if_else_endif() {
let text = "{{ if foo }}Hello!{{ else }}Goodbye!{{ endif }}";
let instructions = compile(text).unwrap();
assert_eq!(4, instructions.len());
assert_eq!(
&Branch(vec![PathStep::Name("foo")], true, 3),
&instructions[0]
);
assert_eq!(&Literal("Hello!"), &instructions[1]);
assert_eq!(&Goto(4), &instructions[2]);
assert_eq!(&Literal("Goodbye!"), &instructions[3]);
}
#[test]
fn test_with() {
let text = "{{ with foo as bar }}Hello!{{ endwith }}";
let instructions = compile(text).unwrap();
assert_eq!(3, instructions.len());
assert_eq!(
&PushNamedContext(vec![PathStep::Name("foo")], "bar"),
&instructions[0]
);
assert_eq!(&Literal("Hello!"), &instructions[1]);
assert_eq!(&PopContext, &instructions[2]);
}
#[test]
fn test_foreach() {
let text = "{{ for foo in bar.baz }}{ foo }{{ endfor }}";
let instructions = compile(text).unwrap();
assert_eq!(5, instructions.len());
assert_eq!(
&PushIterationContext(vec![PathStep::Name("bar"), PathStep::Name("baz")], "foo"),
&instructions[0]
);
assert_eq!(&Iterate(4), &instructions[1]);
assert_eq!(&Value(vec![PathStep::Name("foo")]), &instructions[2]);
assert_eq!(&Goto(1), &instructions[3]);
assert_eq!(&PopContext, &instructions[4]);
}
#[test]
fn test_strip_whitespace_value() {
let text = "Hello, {- name -} , how are you?";
let instructions = compile(text).unwrap();
assert_eq!(3, instructions.len());
assert_eq!(&Literal("Hello,"), &instructions[0]);
assert_eq!(&Value(vec![PathStep::Name("name")]), &instructions[1]);
assert_eq!(&Literal(", how are you?"), &instructions[2]);
}
#[test]
fn test_strip_whitespace_block() {
let text = "Hello, {{- if name -}} {name} {{- endif -}} , how are you?";
let instructions = compile(text).unwrap();
assert_eq!(6, instructions.len());
assert_eq!(&Literal("Hello,"), &instructions[0]);
assert_eq!(
&Branch(vec![PathStep::Name("name")], true, 5),
&instructions[1]
);
assert_eq!(&Literal(""), &instructions[2]);
assert_eq!(&Value(vec![PathStep::Name("name")]), &instructions[3]);
assert_eq!(&Literal(""), &instructions[4]);
assert_eq!(&Literal(", how are you?"), &instructions[5]);
}
#[test]
fn test_comment() {
let text = "Hello, {# foo bar baz #} there!";
let instructions = compile(text).unwrap();
assert_eq!(2, instructions.len());
assert_eq!(&Literal("Hello, "), &instructions[0]);
assert_eq!(&Literal(" there!"), &instructions[1]);
}
#[test]
fn test_strip_whitespace_comment() {
let text = "Hello, \t\n {#- foo bar baz -#} \t there!";
let instructions = compile(text).unwrap();
assert_eq!(2, instructions.len());
assert_eq!(&Literal("Hello,"), &instructions[0]);
assert_eq!(&Literal("there!"), &instructions[1]);
}
#[test]
fn test_strip_whitespace_followed_by_another_tag() {
let text = "{value -}{value} Hello";
let instructions = compile(text).unwrap();
assert_eq!(3, instructions.len());
assert_eq!(&Value(vec![PathStep::Name("value")]), &instructions[0]);
assert_eq!(&Value(vec![PathStep::Name("value")]), &instructions[1]);
assert_eq!(&Literal(" Hello"), &instructions[2]);
}
#[test]
fn test_call() {
let text = "{{ call my_macro with foo.bar }}";
let instructions = compile(text).unwrap();
assert_eq!(1, instructions.len());
assert_eq!(
&Call(
"my_macro",
vec![PathStep::Name("foo"), PathStep::Name("bar")]
),
&instructions[0]
);
}
#[test]
fn test_curly_brace_escaping() {
let text = "body \\{ \nfont-size: {fontsize} \n}";
let instructions = compile(text).unwrap();
assert_eq!(4, instructions.len());
assert_eq!(&Literal("body "), &instructions[0]);
assert_eq!(&Literal("{ \nfont-size: "), &instructions[1]);
assert_eq!(&Value(vec![PathStep::Name("fontsize")]), &instructions[2]);
assert_eq!(&Literal(" \n}"), &instructions[3]);
}
#[test]
fn test_unclosed_tags() {
let tags = vec![
"{",
"{ foo.bar",
"{ foo.bar\n }",
"{{",
"{{ if foo.bar",
"{{ if foo.bar \n}}",
"{#",
"{# if foo.bar",
"{# if foo.bar \n#}",
];
for tag in tags {
compile(tag).unwrap_err();
}
}
#[test]
fn test_mismatched_blocks() {
let text = "{{ if foo }}{{ with bar }}{{ endif }} {{ endwith }}";
compile(text).unwrap_err();
}
#[test]
fn test_disallows_invalid_keywords() {
let text = "{ @foo }";
compile(text).unwrap_err();
}
#[test]
fn test_diallows_unknown_block_type() {
let text = "{{ foobar }}";
compile(text).unwrap_err();
}
#[test]
fn test_parse_error_line_column_num() {
let text = "\n\n\n{{ foobar }}";
let err = compile(text).unwrap_err();
if let ParseError { line, column, .. } = err {
assert_eq!(4, line);
assert_eq!(3, column);
} else {
panic!("Should have returned a parse error");
}
}
#[test]
fn test_parse_error_on_unclosed_if() {
let text = "{{ if foo }}";
compile(text).unwrap_err();
}
#[test]
fn test_parse_escaped_open_curly_brace() {
let text: &str = r"hello \{world}";
let instructions = compile(text).unwrap();
assert_eq!(2, instructions.len());
assert_eq!(&Literal("hello "), &instructions[0]);
assert_eq!(&Literal("{world}"), &instructions[1]);
}
}