sailfish/sailfish/src/filter.rs

//! Build-in filters

use std::fmt;
use std::ptr;

use crate::{Buffer, Escape, Render, RenderError, RenderOnce};

/// Helper struct for 'display' filter
#[derive(Clone, Copy)]
pub struct Display<T>(T);

impl<T: fmt::Display> Render for Display<T> {
    fn render(&self, b: &mut Buffer) -> Result<(), RenderError> {
        use fmt::Write;

        write!(b, "{}", self.0).map_err(RenderError::from)
    }
}

/// render using `std::fmt::Display` trait
///
/// # Examples
///
/// ```text
/// filename: <%\html filename.display() | disp %>
/// ```
#[inline]
pub fn disp<T: fmt::Display>(expr: T) -> Display<T> {
    Display(expr)
}

/// Helper struct for 'dbg' filter
#[derive(Clone, Copy)]
pub struct Debug<T>(T);

impl<T: fmt::Debug> Render for Debug<T> {
    fn render(&self, b: &mut Buffer) -> Result<(), RenderError> {
        use fmt::Write;

        write!(b, "{:?}", self.0).map_err(RenderError::from)
    }
}

/// render using `std::fmt::Debug` trait
///
/// # Examples
///
/// The following examples produce exactly same results, but former is faster
///
/// ```text
/// table content: <%\html table | dbg %>
/// ```
///
/// ```text
/// table content: <%\html format!("{:?}", table) %>
/// ```
#[inline]
pub fn dbg<T: fmt::Debug>(expr: T) -> Debug<T> {
    Debug(expr)
}

/// Helper struct for 'upper' filter
#[derive(Clone, Copy)]
pub struct Upper<T>(T);

impl<T: RenderOnce> RenderOnce for Upper<T> {
    #[inline]
    fn render_once(self, b: &mut Buffer) -> Result<(), RenderError> {
        let mut tmp = Buffer::new();
        self.0.render_once(&mut tmp)?;
        // Estimate assuming ASCII and non-convertible UTF-8 are the most common.
        b.reserve(tmp.len());
        for c in tmp.as_str().chars().flat_map(|c| c.to_uppercase()) {
            b.push(c);
        }
        Ok(())
    }

    #[inline]
    fn render_once_escaped<E: Escape>(
        self,
        b: &mut Buffer,
        e: &E,
    ) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once(b)?;
        let mut tmp = Buffer::new();
        let s = &b.as_str()[old_len..];
        // Estimate assuming ASCII and non-convertible UTF-8 are the most common.
        tmp.reserve(s.len());
        for c in s.chars().flat_map(|c| c.to_uppercase()) {
            tmp.push(c);
        }
        unsafe { b._set_len(old_len) };
        e.escape_to_buf(b, tmp.as_str());
        Ok(())
    }
}

/// convert the rendered contents to uppercase
///
/// # Examples
///
/// ```text
/// <%\html "tschüß" | upper %>
/// ```
///
/// result:
///
/// ```text
/// TSCHÜSS
/// ```
#[inline]
pub fn upper<T: RenderOnce>(expr: T) -> Upper<T> {
    Upper(expr)
}

/// Helper struct for 'lower' filter
#[derive(Clone, Copy)]
pub struct Lower<T>(T);

impl<T: RenderOnce> RenderOnce for Lower<T> {
    #[inline]
    fn render_once(self, b: &mut Buffer) -> Result<(), RenderError> {
        let mut tmp = Buffer::new();
        self.0.render_once(&mut tmp)?;
        // Estimate assuming ASCII and non-convertible UTF-8 are the most common.
        b.reserve(tmp.len());
        let old_len = b.len();
        for c in tmp.as_str().chars() {
            // See comments in str::to_lowercase
            if c == 'Σ' {
                // This is very inefficient, but it seems unlikely to be worth the effort
                // duplicating what already exists in the standard library for unicode support.
                let lower = tmp.as_str().to_lowercase();
                unsafe { b._set_len(old_len) };
                b.push_str(&lower);
                return Ok(());
            }
            for c in c.to_lowercase() {
                b.push(c);
            }
        }
        Ok(())
    }

    #[inline]
    fn render_once_escaped<E: Escape>(
        self,
        b: &mut Buffer,
        e: &E,
    ) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once(b)?;
        let mut tmp = Buffer::new();
        let s = &b.as_str()[old_len..];
        // Estimate assuming ASCII and non-convertible UTF-8 are the most common.
        tmp.reserve(s.len());
        for c in s.chars() {
            // see comments in str::to_lowercase
            if c == 'Σ' {
                let lower = s.to_lowercase();
                unsafe { b._set_len(old_len) };
                e.escape_to_buf(b, &lower);
                return Ok(());
            }
            for c in c.to_lowercase() {
                tmp.push(c);
            }
        }
        unsafe { b._set_len(old_len) };
        e.escape_to_buf(b, tmp.as_str());
        Ok(())
    }
}

/// convert the rendered contents to lowercase
///
/// # Examples
///
/// ```text
/// <%\html "ὈΔΥΣΣΕΎΣ" | lower %>
/// ```
///
/// result:
///
/// ```text
/// ὀδυσσεύς
/// ```
#[inline]
pub fn lower<T: RenderOnce>(expr: T) -> Lower<T> {
    Lower(expr)
}

/// Helper struct for 'trim' filter
#[derive(Clone, Copy)]
pub struct Trim<T>(T);

impl<T: RenderOnce> RenderOnce for Trim<T> {
    #[inline]
    fn render_once(self, b: &mut Buffer) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once(b)?;
        trim_impl(b, old_len)
    }

    #[inline]
    fn render_once_escaped<E: Escape>(
        self,
        b: &mut Buffer,
        e: &E,
    ) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once_escaped(b, e)?;
        trim_impl(b, old_len)
    }
}

fn trim_impl(b: &mut Buffer, old_len: usize) -> Result<(), RenderError> {
    let new_contents = b.as_str().get(old_len..).ok_or(RenderError::BufSize)?;

    let trimmed = new_contents.trim();
    let trimmed_len = trimmed.len();

    if new_contents.len() != trimmed_len {
        // performs inplace trimming

        if new_contents.as_ptr() != trimmed.as_ptr() {
            debug_assert!(new_contents.as_ptr() < trimmed.as_ptr());
            let offset = trimmed.as_ptr() as usize - new_contents.as_ptr() as usize;
            unsafe {
                ptr::copy(
                    b.as_mut_ptr().add(old_len + offset),
                    b.as_mut_ptr().add(old_len),
                    trimmed_len,
                );
            }
        }

        debug_assert!(b.capacity() >= old_len + trimmed_len);

        // SAFETY: `new_contents.len() = b.len() - old_len` and
        // `trimmed_len < new_contents.len()`, so `old_len + trimmed_len < b.len()`.
        unsafe {
            b._set_len(old_len + trimmed_len);
        }
    }

    Ok(())
}

/// Remove leading and trailing writespaces from rendered results
///
/// # Examples
///
/// ```text
/// <%\html " Hello world\n" | trim %>
/// ```
///
/// result:
///
/// ```text
/// Hello world
/// ```
#[inline]
pub fn trim<T: RenderOnce>(expr: T) -> Trim<T> {
    Trim(expr)
}

/// Helper struct for 'truncate' filter
#[derive(Clone, Copy)]
pub struct Truncate<T>(T, usize);

impl<T: RenderOnce> RenderOnce for Truncate<T> {
    #[inline]
    fn render_once(self, b: &mut Buffer) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once(b)?;
        truncate_impl(b, old_len, self.1)
    }

    #[inline]
    fn render_once_escaped<E: Escape>(
        self,
        b: &mut Buffer,
        e: &E,
    ) -> Result<(), RenderError> {
        let old_len = b.len();
        self.0.render_once_escaped(b, e)?;
        truncate_impl(b, old_len, self.1)
    }
}

#[cfg_attr(feature = "perf-inline", inline)]
fn truncate_impl(
    b: &mut Buffer,
    old_len: usize,
    limit: usize,
) -> Result<(), RenderError> {
    let new_contents = b.as_str().get(old_len..).ok_or(RenderError::BufSize)?;
    if new_contents.len() > limit {
        if let Some(idx) = new_contents.char_indices().nth(limit).map(|(i, _)| i) {
            unsafe { b._set_len(old_len + idx) };
            b.push_str("...");
        }
    }

    Ok(())
}

/// Limit length of rendered contents, appends '...' if truncated
///
/// # Examples
///
/// The following example renders the first 20 characters of `message`
///
/// ```test
/// <%\html "Hello, world!" | truncate(5) %>
/// ```
///
/// result:
///
/// ```text
/// Hello...
/// ```
#[inline]
pub fn truncate<T: RenderOnce>(expr: T, limit: usize) -> Truncate<T> {
    Truncate(expr, limit)
}

cfg_json! {
    /// Helper struct for 'json' filter
    #[derive(Clone, Copy)]
    pub struct Json<T>(T);

    impl<T: serde::Serialize> Render for Json<T> {
        #[inline]
        fn render(&self, b: &mut Buffer) -> Result<(), RenderError> {
            struct Writer<'a>(&'a mut Buffer);

            impl<'a> std::io::Write for Writer<'a> {
                #[inline]
                fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
                    let buf = unsafe { std::str::from_utf8_unchecked(buf) };
                    self.0.push_str(buf);
                    Ok(buf.len())
                }

                #[inline]
                fn write_all(&mut self, buf: &[u8]) -> std::io::Result<()> {
                    self.write(buf).map(|_| {})
                }

                #[inline]
                fn flush(&mut self) -> std::io::Result<()> {
                    Ok(())
                }
            }

            serde_json::to_writer(Writer(b), &self.0)
                .map_err(|e| RenderError::new(&e.to_string()))
        }

        #[inline]
        fn render_escaped<E: Escape>(
            &self,
            b: &mut Buffer,
            e: &E,
        ) -> Result<(), RenderError> {
            struct Writer<'a, E: Escape>(&'a mut Buffer, &'a E);

            impl<'a, E: Escape> std::io::Write for Writer<'a, E> {
                #[inline]
                fn write(&mut self, buf: &[u8]) -> std::io::Result<usize> {
                    let buf = unsafe { std::str::from_utf8_unchecked(buf) };
                    self.1.escape_to_buf(self.0, buf);
                    Ok(buf.len())
                }

                #[inline]
                fn write_all(&mut self, buf: &[u8]) -> std::io::Result<()> {
                    self.write(buf).map(|_| {})
                }

                #[inline]
                fn flush(&mut self) -> std::io::Result<()> {
                    Ok(())
                }
            }

            serde_json::to_writer(Writer(b, e), &self.0)
                .map_err(|e| RenderError::new(&e.to_string()))
        }
    }

    /// Serialize the given data structure as JSON into the buffer
    ///
    /// # Examples
    ///
    /// ```text
    /// {
    ///     "name": "JSON example",
    ///     "data": <%- data | json %>
    /// }
    /// ```
    #[inline]
    pub fn json<T: serde::Serialize>(expr: T) -> Json<T> {
        Json(expr)
    }
}

#[cfg(test)]
mod tests {
    use crate::EscapeHtml;

    use super::*;

    fn assert_render<T: RenderOnce>(expr: T, expected: &str) {
        let mut buf = Buffer::new();
        RenderOnce::render_once(expr, &mut buf).unwrap();
        assert_eq!(buf.as_str(), expected);
    }

    fn assert_render_escaped<T: RenderOnce>(expr: T, expected: &str) {
        let mut buf = Buffer::new();
        RenderOnce::render_once_escaped(expr, &mut buf, &EscapeHtml).unwrap();
        assert_eq!(buf.as_str(), expected);
    }

    #[test]
    fn test_lower() {
        assert_render(lower(""), "");
        assert_render_escaped(lower(""), "");

        assert_render(lower("lorem ipsum"), "lorem ipsum");
        assert_render(lower("LOREM IPSUM"), "lorem ipsum");

        assert_render_escaped(lower("hElLo, WOrLd!"), "hello, world!");
        assert_render_escaped(lower("hElLo, WOrLd!"), "hello, world!");

        assert_render_escaped(lower("<h1>TITLE</h1>"), "&lt;h1&gt;title&lt;/h1&gt;");
        assert_render_escaped(lower("<<&\"\">>"), "&lt;&lt;&amp;&quot;&quot;&gt;&gt;");

        // non-ascii
        assert_render(lower("aBcＡｂｃ"), "abcａｂｃ");
        assert_render(lower("ὈΔΥΣΣΕΎΣ"), "ὀδυσσεύς");
    }

    #[test]
    fn test_upper() {
        assert_render(upper(""), "");
        assert_render_escaped(upper(""), "");

        assert_render(upper("lorem ipsum"), "LOREM IPSUM");
        assert_render(upper("LOREM IPSUM"), "LOREM IPSUM");

        assert_render(upper("hElLo, WOrLd!"), "HELLO, WORLD!");
        assert_render(upper("hElLo, WOrLd!"), "HELLO, WORLD!");

        // non-ascii
        assert_render(upper("aBcＡｂｃ"), "ABCＡＢＣ");
        assert_render(upper("tschüß"), "TSCHÜSS");
    }

    #[test]
    fn test_trim() {
        assert_render(trim(""), "");
        assert_render_escaped(trim(""), "");

        assert_render(trim("\n \t\r\x0C"), "");

        assert_render(trim("hello world!"), "hello world!");
        assert_render(trim("hello world!\n"), "hello world!");
        assert_render(trim("\thello world!"), "hello world!");
        assert_render(trim("\thello world!\r\n"), "hello world!");

        assert_render_escaped(trim(" <html>  "), "&lt;html&gt;");
        assert_render_escaped(lower("<<&\"\">>"), "&lt;&lt;&amp;&quot;&quot;&gt;&gt;");

        // non-ascii whitespace
        assert_render(trim("\u{A0}空白\u{3000}\u{205F}"), "空白");
    }

    #[test]
    fn test_truncate() {
        assert_render(truncate("", 0), "");
        assert_render(truncate("", 5), "");

        assert_render(truncate("apple ", 0), "...");
        assert_render(truncate("apple ", 1), "a...");
        assert_render(truncate("apple ", 2), "ap...");
        assert_render(truncate("apple ", 3), "app...");
        assert_render(truncate("apple ", 4), "appl...");
        assert_render(truncate("apple ", 5), "apple...");
        assert_render(truncate("apple ", 6), "apple ");
        assert_render(truncate("apple ", 7), "apple ");

        assert_render(truncate(std::f64::consts::PI, 10), "3.14159265...");
        assert_render(truncate(std::f64::consts::PI, 20), "3.141592653589793");

        assert_render_escaped(truncate("foo<br>bar", 10), "foo&lt;br&...");
        assert_render_escaped(truncate("foo<br>bar", 20), "foo&lt;br&gt;bar");

        // non-ascii
        assert_render(truncate("魑魅魍魎", 0), "...");
        assert_render(truncate("魑魅魍魎", 1), "魑...");
        assert_render(truncate("魑魅魍魎", 2), "魑魅...");
        assert_render(truncate("魑魅魍魎", 3), "魑魅魍...");
        assert_render(truncate("魑魅魍魎", 4), "魑魅魍魎");
        assert_render(truncate("魑魅魍魎", 5), "魑魅魍魎");
    }

    #[cfg(feature = "json")]
    #[test]
    fn test_json() {
        assert_render(json(""), "\"\"");
        assert_render(json(serde_json::json!({})), "{}");

        assert_render_escaped(json(123_i32), "123");
        assert_render_escaped(json("Pokémon"), "&quot;Pokémon&quot;");
    }

    #[test]
    fn compine() {
        assert_render(
            lower(upper("Li Europan lingues es membres del sam familie.")),
            "li europan lingues es membres del sam familie.",
        );
        assert_render(lower(lower("ハートのＡ")), "ハートのａ");
        assert_render(upper(upper("ハートのＡ")), "ハートのＡ");

        assert_render(truncate(trim("\t起来！\r\n"), 1), "起...");
        assert_render(truncate(trim("\t起来！\r\n"), 3), "起来！");

        assert_render(truncate(lower("Was möchtest du?"), 10), "was möchte...");
        assert_render(truncate(upper("Was möchtest du?"), 10), "WAS MÖCHTE...");
    }
}