HashMap<UniCase<Cow<str>>>生命周期问题的修复方案咨询

问题背景 #

UniCase未实现Borrow trait，导致无法在不分配String的情况下用UniCase键索引HashMap。官方给出的临时方案是使用HashMap<UniCase<Cow<str>>>，但以下代码编译失败：

use std::borrow::Cow;
use std::collections::HashMap;
use std::io::{BufRead, Result};

use unicase::UniCase;

fn rm_reader_keys<R: BufRead>(
    reader: R, dict: &mut HashMap<UniCase<Cow<str>>, String>
) -> Result<()> {
    for ln in reader.lines() {
        let line: String = ln?;
        if let Some((key, _)) = line.rsplit_once(' ') {
            dict.remove(&UniCase::unicode(Cow::Borrowed(key)));
        }
    }
    Ok(())
}

编译报错信息 #

error[E0597]: `line` does not live long enough
  --> src/main.rs:44:33
   |
40 |     reader: R, dict: &mut HashMap<UniCase<Cow<str>>, std::string::String>
   |                ---- has type `&mut HashMap<UniCase<Cow<'1, str>>, std::string::String>`
...
43 |         let line: String = ln?;
   |             ---- binding `line` declared here
44 |         if let Some((key, _)) = line.rsplit_once(' ') {
   |                                 ^^^^^^^^^^^^^^^^^^^^^ borrowed value does not live long enough
45 |             dict.remove(&UniCase::unicode(Cow::Borrowed(key)));
   |             -------------------------------------------------- argument requires that `line` is borrowed for `'1`
46 |         }
47 |     }
   |     - `line` dropped here while still borrowed

报错原因 #

dict的生命周期'1要求传入的键引用必须存活到'1的整个周期，但line是循环内的局部变量，每次循环结束就会被销毁，无法满足这个生命周期约束，导致编译报错。

可行修复方法 #

方法一：将Cow转为Owned（分配String） #

把key转换为Cow::Owned类型，让UniCase内部持有独立的String，不再依赖line的生命周期：

use std::borrow::Cow;
use std::collections::HashMap;
use std::io::{BufRead, Result};

use unicase::UniCase;

fn rm_reader_keys<R: BufRead>(
    reader: R, dict: &mut HashMap<UniCase<Cow<str>>, String>
) -> Result<()> {
    for ln in reader.lines() {
        let line: String = ln?;
        if let Some((key, _)) = line.rsplit_once(' ') {
            dict.remove(&UniCase::unicode(Cow::Owned(key.to_string())));
        }
    }
    Ok(())
}

方法二：遍历HashMap匹配键（避免分配） #

利用UniCase::as_str方法遍历HashMap，找到大小写匹配的键后再移除，这种方式不需要分配新的String：

use std::borrow::Cow;
use std::collections::HashMap;
use std::io::{BufRead, Result};

use unicase::UniCase;

fn rm_reader_keys<R: BufRead>(
    reader: R, dict: &mut HashMap<UniCase<Cow<str>>, String>
) -> Result<()> {
    for ln in reader.lines() {
        let line: String = ln?;
        if let Some((key, _)) = line.rsplit_once(' ') {
            let target_key = dict.keys()
                .find(|uc| uc.as_str().eq_ignore_ascii_case(key))
                .cloned();
            if let Some(k) = target_key {
                dict.remove(&k);
            }
        }
    }
    Ok(())
}

方法三：改用UniCase作为键类型 #

如果可以接受初始就使用UniCase<String>作为HashMap的键类型，代码会更简洁，也不会有生命周期问题：

use std::collections::HashMap;
use std::io::{BufRead, Result};

use unicase::UniCase;

fn rm_reader_keys<R: BufRead>(
    reader: R, dict: &mut HashMap<UniCase<String>, String>
) -> Result<()> {
    for ln in reader.lines() {
        let line: String = ln?;
        if let Some((key, _)) = line.rsplit_once(' ') {
            dict.remove(&UniCase::unicode(key.to_string()));
        }
    }
    Ok(())
}

内容的提问来源于stack exchange，提问作者Nil Admirari