Regex translation: `Regex.Replace` substitutions are passed through, `Singleline | Multiline` gets the wrong flag, and `[^x]` never matches a newline

未关闭
  • 哪里不对？
#4,094 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

维护者通常 1 天内回复

还没有人认领这个 Issue。

评估

难度
3/5
预计耗时
1-2 天
新手友好度
55/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
csharp, postgresql, sql
领域
database

调研方向

Start in src/EFCore.PG/Query/ExpressionTranslators/Internal/NpgsqlRegexTranslator.cs (lines 111-119, 166-170) and src/EFCore.PG/Query/Internal/NpgsqlQuerySqlGenerator.cs (lines 1022-1030) to see how regex flags and substitutions are emitted. Run the repro DbContext from the issue to observe mismatched SQL. Done means replacements use \1-style substitutions and the w flag for Singleline|Multiline, or a docs note covers gaps.

由索引模型根据 Issue 内容生成。

描述

Environment
PostgreSQL 18.6 (postgres:18), session TimeZone Etc/UTC
.NET 10 runtime 10.0.11, Npgsql.EntityFrameworkCore.PostgreSQL 10.0.3, EF Core 10.0.4
.NET 11 runtime 11.0.0-rc.2, Npgsql.EntityFrameworkCore.PostgreSQL 11.0.0-rc.2, EF Core 11.0.0-rc.2
Also Npgsql 10.0.3 on both lines, macOS arm64

Both lines print the same output. main at 150011d has the same code.

LINQ Input .NET PostgreSQL SQL
Regex.Replace(r.Text, "(a)", "[$1]") abc [a]bc [$1]bc regexp_replace(r."Text", '(a)', '[$1]', 'p')
Regex.IsMatch(r.Text, "a.b", RegexOptions.Singleline | RegexOptions.Multiline) a\nb match no match r."Text" ~ '(?n)a.b'
Regex.IsMatch(r.Text, "a[^x]b") a\nb match no match r."Text" ~ '(?p)a[^x]b'
Repro 
using Microsoft.EntityFrameworkCore;
using Microsoft.Extensions.Logging;
using System.Text.RegularExpressions;

await using var context = new TestContext();
await context.Database.EnsureDeletedAsync();
await context.Database.EnsureCreatedAsync();

context.Rows.AddRange(new Row { Text = "abc" }, new Row { Text = "a\nb" });
await context.SaveChangesAsync();
var rows = await context.Rows.AsNoTracking().OrderBy(r => r.Id).ToListAsync();

// 1. .NET substitution syntax is passed through; PostgreSQL only knows \1..\9 and \&
Console.WriteLine($"Regex.Replace(\"abc\", \"(a)\", \"[$1]\"): .NET {Regex.Replace(rows[0].Text, "(a)", "[$1]")}, database {await context.Rows.Where(r => r.Id == 1).Select(r => Regex.Replace(r.Text, "(a)", "[$1]")).SingleAsync()}");

// 2. Singleline makes . match \n, also together with Multiline
Console.WriteLine($"IsMatch(\"a\\nb\", \"a.b\", Singleline | Multiline): .NET {rows.Count(r => Regex.IsMatch(r.Text, "a.b", RegexOptions.Singleline | RegexOptions.Multiline))}, database {await context.Rows.CountAsync(r => Regex.IsMatch(r.Text, "a.b", RegexOptions.Singleline | RegexOptions.Multiline))}");

// 3. A negated bracket matches \n in .NET
Console.WriteLine($"IsMatch(\"a\\nb\", \"a[^x]b\"): .NET {rows.Count(r => Regex.IsMatch(r.Text, "a[^x]b"))}, database {await context.Rows.CountAsync(r => Regex.IsMatch(r.Text, "a[^x]b"))}");

public class TestContext : DbContext
{
    public DbSet<Row> Rows => Set<Row>();

    protected override void OnConfiguring(DbContextOptionsBuilder optionsBuilder)
        => optionsBuilder
            .UseNpgsql(Environment.GetEnvironmentVariable("Test__Npgsql__DefaultConnection"), o => o.UseNodaTime())
            .LogTo(Console.WriteLine, LogLevel.Information)
            .EnableSensitiveDataLogging();
}

public class Row
{
    public int Id { get; set; }
    public string Text { get; set; } = "";
}

Output, the same on both lines:

Regex.Replace("abc", "(a)", "[$1]"): .NET [a]bc, database [$1]bc
IsMatch("a\nb", "a.b", Singleline | Multiline): .NET 1, database 0
IsMatch("a\nb", "a[^x]b"): .NET 1, database 0
  • Substitutions are passed through unchanged (NpgsqlRegexTranslator.cs#L111-L119). PostgreSQL expands \1–\9 and \&. .NET uses $1, ${name}, $& and $$, and treats a backslash literally. So $1 comes out as $1, and \1 means different things on the two sides. A constant replacement could be rewritten; a column replacement can't easily be.
  • Singleline | Multiline renders n because the Multiline check comes first (NpgsqlQuerySqlGenerator.cs#L1022-L1030, NpgsqlRegexTranslator.cs#L166-L170). With n, . doesn't match a newline. .NET's Singleline with Multiline corresponds to w: ^ and $ per line, and . matches newlines.
  • The default p flag, which is there so that . doesn't match a newline, also stops a negated bracket like [^x] from matching one. .NET's default lets [^x] match \n. There may be no flag combination that matches .NET exactly, so a docs note may be the realistic fix here. The same goes for other dialect differences, for example \b, which is a backspace in PostgreSQL's ARE (a word boundary there is \y).

Regex.Replace on 10.0.3 and 11.0.0-rc.2 still uses 'p' without g. That is already fixed on main and in hotfix/10.0.4 (#4035).

主要语言
C#
星标
1.8k
派生
268
平均合并
2 小时 35 分钟
30 天内合并 PR
136

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue，再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库，在一个分支上完成修改。
  4. 提交 Pull Request，并在描述里引用这个 Issue 编号。

npgsql/efcore.pg 的其他 Issue

查看 npgsql/efcore.pg 的全部 Issue

相似的 Issue

更多 C# Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。