build.rs

use fs2::FileExt;
use std::io::{Read, Write};
use std::{borrow::Cow, collections::HashMap, io::BufWriter, path::Path};

const GENERATION_PATH: &str = "include/zenoh-gen.h";
const SPLITGUIDE_PATH: &str = "splitguide.yaml";
const HEADER: &str = r"//
// Copyright (c) 2022 ZettaScale Technology
//
// This program and the accompanying materials are made available under the
// terms of the Eclipse Public License 2.0 which is available at
// http://www.eclipse.org/legal/epl-2.0, or the Apache License, Version 2.0
// which is available at https://www.apache.org/licenses/LICENSE-2.0.
//
// SPDX-License-Identifier: EPL-2.0 OR Apache-2.0
//
// Contributors:
//   ZettaScale Zenoh Team, <zenoh@zettascale.tech>
//
#ifdef DOCS
#define ALIGN(n)
#define ZENOHC_API
#endif
";

fn main() {
    cbindgen::generate(std::env::var("CARGO_MANIFEST_DIR").unwrap())
        .expect("Unable to generate bindings")
        .write_to_file(GENERATION_PATH);

    configure();
    let split_guide = SplitGuide::from_yaml(SPLITGUIDE_PATH);
    split_bindings(&split_guide).unwrap();
    text_replace(split_guide.rules.iter().map(|(name, _)| name.as_str()));

    println!("cargo:rerun-if-changed=build.rs");
    println!("cargo:rerun-if-changed=src");
    println!("cargo:rerun-if-changed=splitguide.yaml");
    println!("cargo:rerun-if-changed=cbindgen.toml")
}

fn configure() {
    let content = format!(
        r#"#pragma once
#define TARGET_ARCH_{}
"#,
        std::env::var("CARGO_CFG_TARGET_ARCH")
            .unwrap()
            .to_uppercase()
    );

    let mut file = std::fs::File::options()
        .write(true)
        .truncate(true)
        .append(false)
        .create(true)
        .open("include/zenoh_configure.h")
        .unwrap();
    file.lock_exclusive().unwrap();
    file.write_all(content.as_bytes()).unwrap();
    file.unlock().unwrap();
}

fn text_replace<'a>(files: impl Iterator<Item = &'a str>) {
    for name in files {
        let path = format!("include/{}", name);

        // Read content
        let mut file = std::fs::File::options()
            .read(true)
            .create(false)
            .write(false)
            .open(&path)
            .unwrap();
        file.lock_exclusive().unwrap();
        let mut buf = String::new();
        file.read_to_string(&mut buf).unwrap();
        file.unlock().unwrap();

        // Remove _T_ from enum variant name
        let buf = buf.replace("_T_", "_");
        // Replace _t_Tag from union variant name
        let buf = buf.replace("_t_Tag", "_tag_t");
        // Insert `ZENOHC_API` macro before `extern const`.
        // The cbindgen can prefix functions (see `[fn] prefix=...` in cbindgn.toml), but not extern variables.
        // So have to do it here.
        let buf = buf.replace("extern const", "ZENOHC_API extern const");

        // Overwrite content
        let mut file = std::fs::File::options()
            .read(false)
            .create(false)
            .write(true)
            .truncate(true)
            .open(&path)
            .unwrap();
        file.lock_exclusive().unwrap();
        file.write_all(buf.as_bytes()).unwrap();
        file.unlock().unwrap();
    }
}

fn split_bindings(split_guide: &SplitGuide) -> Result<(), String> {
    let bindings = std::fs::read_to_string(GENERATION_PATH).unwrap();
    let mut files = split_guide
        .rules
        .iter()
        .map(|(name, _)| {
            let file = std::fs::File::options()
                .write(true)
                .truncate(true)
                .append(false)
                .create(true)
                .open(format!("include/{}", name))
                .unwrap();
            file.lock_exclusive().unwrap();
            file.set_len(0).unwrap();
            (name.as_str(), BufWriter::new(file))
        })
        .collect::<HashMap<_, _>>();
    for file in files.values_mut() {
        file.write_all(HEADER.as_bytes())
            .map_err(|e| e.to_string())?;
    }
    let mut records = group_tokens(Tokenizer {
        filename: GENERATION_PATH,
        inner: &bindings,
    })?;
    for id in split_guide.requested_ids() {
        if !records.iter().any(|r| r.contains_id(id)) {
            return Err(format!(
                "{} not found (requested explicitly by splitguide.yaml)",
                id,
            ));
        }
    }
    for record in &mut records {
        let appropriate_files = split_guide.appropriate_files(record);
        for file in appropriate_files {
            let writer = files.get_mut(file).unwrap();
            record.used = true;
            write!(writer, "{}", &record).unwrap();
        }
    }
    for record in &records {
        record.is_used()?;
    }
    for (_, file) in files {
        file.into_inner().unwrap().unlock().unwrap();
    }
    std::fs::remove_file(GENERATION_PATH).unwrap();
    Ok(())
}

enum SplitRule {
    Brand(RecordType),
    Exclusive(String),
    Shared(String),
}
struct SplitGuide {
    rules: Vec<(String, Vec<SplitRule>)>,
}
impl SplitGuide {
    fn from_yaml<P: AsRef<Path>>(path: P) -> Self {
        let map: HashMap<String, Vec<String>> =
            serde_yaml::from_reader(std::fs::File::open(path).unwrap()).unwrap();
        SplitGuide {
            rules: map
                .into_iter()
                .map(|(name, rules)| {
                    (
                        name,
                        rules
                            .into_iter()
                            .map(|mut s| match s.as_str() {
                                ":functions" => SplitRule::Brand(RecordType::Function),
                                ":typedefs" => SplitRule::Brand(RecordType::Typedef),
                                ":includes" => SplitRule::Brand(RecordType::PreprInclude),
                                ":defines" => SplitRule::Brand(RecordType::PreprDefine),
                                ":const" => SplitRule::Brand(RecordType::Const),
                                ":multiples" => SplitRule::Brand(RecordType::Multiple),
                                _ if s.ends_with('!') => {
                                    s.pop();
                                    SplitRule::Exclusive(s)
                                }
                                _ => SplitRule::Shared(s),
                            })
                            .collect(),
                    )
                })
                .collect(),
        }
    }
    fn appropriate_files(&self, record: &Record) -> Vec<&str> {
        let mut shared = Vec::new();
        let mut exclusives = Vec::new();
        for (file, rules) in &self.rules {
            for rule in rules {
                match rule {
                    SplitRule::Brand(brand) if *brand == record.rt => shared.push(file.as_str()),
                    SplitRule::Exclusive(id) if record.contains_id(id) => {
                        exclusives.push(file.as_str())
                    }
                    SplitRule::Shared(id) if record.contains_id(id) => shared.push(file.as_str()),
                    _ => {}
                }
            }
        }
        if exclusives.is_empty() {
            shared
        } else {
            exclusives
        }
    }
    fn requested_ids(&self) -> impl Iterator<Item = &str> {
        self.rules.iter().flat_map(|(_, rules)| {
            rules.iter().filter_map(|r| match r {
                SplitRule::Brand(_) => None,
                SplitRule::Exclusive(s) | SplitRule::Shared(s) => Some(s.as_str()),
            })
        })
    }
}

fn group_tokens(stream: Tokenizer) -> Result<Vec<Record>, String> {
    let mut records = Vec::new();
    let mut record_collect = Record::new();
    for token in stream {
        record_collect.add_token(token)?;
        if record_collect.is_ready() {
            let mut record = Record::new();
            std::mem::swap(&mut record_collect, &mut record);
            records.push(record);
        }
    }
    records.push(record_collect);
    Ok(records)
}

#[derive(Copy, Clone, PartialEq, Debug)]
enum RecordType {
    Empty,
    Multiple,
    PrivateToken,
    Typedef,
    Function,
    Const,
    PreprDefine,
    PreprInclude,
}

impl RecordType {
    fn update(&mut self, rt: RecordType) {
        match *self {
            RecordType::Empty => *self = rt,
            RecordType::Multiple => {}
            _ => *self = RecordType::Multiple,
        }
    }
}

struct Record<'a> {
    used: bool,
    rt: RecordType,
    nesting: i32,
    ids: Vec<Cow<'a, str>>,
    tokens: Vec<Token<'a>>,
}

impl<'a> Record<'a> {
    fn new() -> Self {
        Self {
            used: false,
            rt: RecordType::Empty,
            nesting: 0,
            ids: Vec::new(),
            tokens: Vec::new(),
        }
    }

    fn is_used(&self) -> Result<(), String> {
        if self.used || self.rt == RecordType::Empty || self.rt == RecordType::PrivateToken {
            Ok(())
        } else {
            let token_ids = self.tokens.iter().map(|t| t.id).collect::<Vec<_>>();
            Err(format!("Unused {:?} record: {:?}", self.rt, token_ids))
        }
    }

    fn is_ready(&self) -> bool {
        self.nesting == 0 && self.rt != RecordType::Empty
    }

    fn contains_id(&self, id: &str) -> bool {
        self.ids.iter().any(|v| v == id)
    }

    fn push_token(&mut self, token: Token<'a>) {
        self.tokens.push(token);
    }

    fn push_record_type_token(&mut self, token: Token<'a>, rt: RecordType) {
        self.rt.update(rt);
        if !token.id.is_empty() {
            self.ids.push(token.id.into());
        }
        self.push_token(token)
    }

    fn push_prepr_if(&mut self, token: Token<'a>) {
        self.nesting += 1;
        self.push_token(token)
    }

    fn push_prepr_endif(&mut self, token: Token<'a>) -> Result<(), String> {
        self.nesting -= 1;
        if self.nesting < 0 {
            return Err("unmatched #endif".into());
        }
        self.push_token(token);
        Ok(())
    }

    fn add_token(&mut self, token: Token<'a>) -> Result<(), String> {
        match token.tt {
            TokenType::Comment => self.push_token(token),
            TokenType::Typedef => self.push_record_type_token(token, RecordType::Typedef),
            TokenType::Function => self.push_record_type_token(token, RecordType::Function),
            TokenType::Const => self.push_record_type_token(token, RecordType::Const),
            TokenType::PrivateToken => self.push_record_type_token(token, RecordType::PrivateToken),
            TokenType::PreprDefine => self.push_record_type_token(token, RecordType::PreprDefine),
            TokenType::PreprInclude => self.push_record_type_token(token, RecordType::PreprInclude),
            TokenType::PreprIf => self.push_prepr_if(token),
            TokenType::PreprElse => self.push_token(token),
            TokenType::PreprEndif => self.push_prepr_endif(token)?,
            TokenType::Whitespace => self.push_token(token),
        }
        Ok(())
    }
}

// Print all comments first, skip whitespaces
impl<'a> std::fmt::Display for Record<'a> {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        self.tokens
            .iter()
            .filter(|t| t.tt == TokenType::Comment)
            .map(|t| t.fmt(f))
            .find(|r| r.is_err())
            .unwrap_or(Ok(()))?;
        self.tokens
            .iter()
            .filter(|t| t.tt != TokenType::Comment && t.tt != TokenType::Whitespace)
            .map(|t| t.fmt(f))
            .find(|r| r.is_err())
            .unwrap_or(Ok(()))?;
        Ok(())
    }
}

#[derive(Debug, Clone, Copy, PartialEq, Eq)]
enum TokenType {
    Comment,
    Typedef,
    Function,
    Const,
    PrivateToken,
    PreprDefine,
    PreprInclude,
    PreprIf,
    PreprElse,
    PreprEndif,
    Whitespace,
}
#[derive(Debug, Clone, PartialEq, Eq)]
struct Token<'a> {
    tt: TokenType,
    id: &'a str,
    span: Cow<'a, str>,
}
impl<'a> std::fmt::Display for Token<'a> {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        // f.write_str(format!("{:?} [", self.tt).as_str())?;
        f.write_str(&self.span)?;
        // f.write_str("]")?;

        // Each token is finalized with endline character on output
        f.write_str("\n")?;
        Ok(())
    }
}
impl<'a> Token<'a> {
    fn new<I: Into<Cow<'a, str>>>(tt: TokenType, id: &'a str, span: I) -> Self {
        Token {
            tt,
            id,
            span: span.into(),
        }
    }

    fn next(s: &'a str) -> Option<Self> {
        Self::whitespace(s)
            .or_else(|| Self::comment(s))
            .or_else(|| Self::prepr_endif(s))
            .or_else(|| Self::prepr_include(s))
            .or_else(|| Self::prepr_define(s))
            .or_else(|| Self::prepr_if(s))
            .or_else(|| Self::prepr_else(s))
            .or_else(|| Self::typedef(s))
            .or_else(|| Self::r#const(s))
            .or_else(|| Self::function(s))
    }

    //
    // Each token is consumed without end of line characters.
    // When performing output of tokens, endline is added to each token.
    // This guarantees that tokens can be shuffled as necessary
    //
    fn typedef(s: &'a str) -> Option<Self> {
        if s.starts_with("typedef") {
            let mut len = 0;
            let mut accolades = 0;
            for c in s.chars() {
                len += c.len_utf8();
                if c == '{' {
                    accolades += 1;
                } else if c == '}' {
                    accolades -= 1;
                } else if c == ';' && accolades == 0 {
                    break;
                }
            }
            let span = &s[..len];
            let id_len = span
                .chars()
                .rev()
                .skip(1)
                .take_while(|&c| c.is_alphanumeric() || c == '_')
                .fold(1, |acc, c| acc + c.len_utf8());
            let id = &span[(span.len() - id_len)..(span.len() - 1)];
            Some(Token::new(
                if id.starts_with('_') {
                    TokenType::PrivateToken
                } else {
                    TokenType::Typedef
                },
                id,
                span,
            ))
        } else {
            None
        }
    }

    fn function(s: &'a str) -> Option<Self> {
        s.until_incl(";").and_then(|span| {
            span.contains('(').then(|| {
                let mut iter = span.chars().rev();
                let id_end = iter
                    .by_ref()
                    .take_while(|&c| c != '(')
                    .fold(1, |acc, c| acc + c.len_utf8());
                let id_len = iter
                    .take_while(|&c| c.is_alphanumeric() || c == '_')
                    .fold(0, |acc, c| acc + c.len_utf8());
                let id = &span[(span.len() - id_end - id_len)..(span.len() - id_end)];
                Token::new(TokenType::Function, id, span)
            })
        })
    }

    fn r#const(s: &'a str) -> Option<Self> {
        s.until_incl(";").and_then(|span| {
            span.contains("const").then(|| {
                let id_len = span
                    .chars()
                    .rev()
                    .skip(1)
                    .take_while(|&c| c.is_alphanumeric() || c == '_')
                    .fold(0, |acc, c| acc + c.len_utf8());
                let id = &span[(span.len() - 1 - id_len)..(span.len() - 1)];
                Token::new(TokenType::Const, id, span)
            })
        })
    }

    fn whitespace(s: &'a str) -> Option<Self> {
        let mut len = 0;
        for c in s.chars() {
            if c.is_whitespace() {
                len += c.len_utf8()
            } else {
                break;
            }
        }
        if len > 0 {
            Some(Token::new(TokenType::Whitespace, "", &s[..len]))
        } else {
            None
        }
    }

    fn comment(s: &'a str) -> Option<Self> {
        if s.starts_with("/*") {
            Some(Token::new(
                TokenType::Comment,
                "",
                s.until_incl("*/").unwrap_or(s),
            ))
        } else if s.starts_with("//") {
            Some(Token::new(
                TokenType::Comment,
                "",
                s.until("\n").unwrap_or(s),
            ))
        } else {
            None
        }
    }

    fn prepr_if(s: &'a str) -> Option<Self> {
        if s.starts_with("#if ") || s.starts_with("#ifdef ") || s.starts_with("#ifndef ") {
            let span = s.until("\n").unwrap_or(s);
            Some(Token::new(TokenType::PreprIf, span, span))
        } else {
            None
        }
    }

    fn prepr_define(s: &'a str) -> Option<Self> {
        let start = "#define ";
        s.strip_prefix(start).map(|defined| {
            let span = s.until("\n").unwrap_or(s);
            Token::new(
                if defined.starts_with('_') {
                    TokenType::PrivateToken
                } else {
                    TokenType::PreprDefine
                },
                span[start.len()..].split_whitespace().next().unwrap(),
                span,
            )
        })
    }

    fn prepr_endif(s: &'a str) -> Option<Self> {
        s.starts_with("#endif")
            .then(|| Token::new(TokenType::PreprEndif, "", s.until("\n").unwrap_or(s)))
    }

    fn prepr_else(s: &'a str) -> Option<Self> {
        s.starts_with("#else")
            .then(|| Token::new(TokenType::PreprElse, "", s.until("\n").unwrap_or(s)))
    }

    fn prepr_include(s: &'a str) -> Option<Self> {
        Self::r#include(s, "#include \"", "\"").or_else(|| Self::r#include(s, "#include <", ">"))
    }

    fn r#include(s: &'a str, start: &str, end: &str) -> Option<Self> {
        if s.starts_with(start) {
            let span = s.until_incl(end).expect("detected unterminated #include");
            Some(Token::new(
                TokenType::PreprInclude,
                &span[start.len()..(span.len() - end.len())],
                span,
            ))
        } else {
            None
        }
    }
}
trait Until: Sized {
    fn until(self, pattern: &str) -> Option<Self>;
    fn until_incl(self, pattern: &str) -> Option<Self>;
}
impl Until for &str {
    fn until(self, pattern: &str) -> Option<Self> {
        self.find(pattern).map(|l| &self[..l])
    }
    fn until_incl(self, pattern: &str) -> Option<Self> {
        self.find(pattern).map(|l| &self[..(l + pattern.len())])
    }
}
struct Tokenizer<'a> {
    filename: &'a str,
    inner: &'a str,
}
impl<'a> Iterator for Tokenizer<'a> {
    type Item = Token<'a>;

    fn next(&mut self) -> Option<Self::Item> {
        if self.inner.is_empty() {
            None
        } else {
            let result = Token::next(self.inner);
            if let Some(result) = &result {
                self.inner = &self.inner[result.span.len()..];
            } else {
                panic!(
                    "Couldn't parse C file {}, stopped at: {}",
                    self.filename,
                    self.inner.lines().next().unwrap()
                )
            }
            result
        }
    }
}