PostgreSQL's ltree extension gives you materialised paths with database-native superpowers: GiST indexes, specialised operators like @> and <@, and powerful pattern matching. If you're committed to PostgreSQL and want the best hierarchy query performance, ltree is hard to beat.

Good news: The Npgsql EF Core provider does support LINQ translations for ltree operations via the LTree type. You can use methods like IsAncestorOf(), IsDescendantOf(), and MatchesLQuery() directly in LINQ queries. However, EF Core does not yet support recursive CTEs, so you'll need raw SQL for operations that require them (like building complete subtree results with calculated depths).

Thanks to Shay Rojansky for pointing out the LINQ translation support!

Series Navigation

• Part 1: Overview - Introduction and comparison
• Part 1.1: Adjacency List
• Part 1.2: Closure Table
• Part 1.3: Materialised Path
• Part 1.4: Nested Sets
• Part 1.5: ltree (this article)

What is ltree?

ltree is a PostgreSQL extension that provides a native data type for hierarchical label paths. Think of it as Materialised Path with superpowers - the database understands the structure and provides optimised operators, functions, and GiST index support.

Instead of treating the path as a dumb string and using LIKE queries, PostgreSQL can:

• Use specialised operators (@> for "is ancestor of", <@ for "is descendant of")
• Apply GiST indexes for efficient hierarchy queries
• Match patterns with wildcards (Top.*.Europe)
• Perform set operations on paths

Key insight: ltree is the best of both worlds - the simplicity of materialised paths with database-native optimisation. The trade-off is PostgreSQL lock-in, and while many ltree operations work via LINQ, recursive CTEs still require raw SQL.

ltree Path Format

Paths in ltree use periods as separators and alphanumeric labels:

Top.Countries.Europe.UK
Top.Countries.Asia.Japan.Tokyo
Top.Products.Electronics.Computers.Laptops

Rules:

• Labels can contain letters, digits, and underscores
• Labels are case-sensitive
• Maximum label length is 256 characters
• Maximum path length is 65535 labels

For comment systems, we'd use IDs as labels: 1.3.7 meaning "comment 7 under comment 3 under comment 1".

Setting Up ltree

First, enable the extension (requires database superuser privileges):

CREATE EXTENSION IF NOT EXISTS ltree;

Or via EF Core migration:

protected override void Up(MigrationBuilder migrationBuilder)
{
    migrationBuilder.Sql("CREATE EXTENSION IF NOT EXISTS ltree");
}

Entity Definition

The Npgsql provider includes an LTree type that maps directly to PostgreSQL's ltree and provides LINQ-translatable methods:

using Microsoft.EntityFrameworkCore;

public class Comment
{
    public int Id { get; set; }
    public string Content { get; set; } = string.Empty;
    public string Author { get; set; } = string.Empty;
    public DateTime CreatedAt { get; set; }

    public int PostId { get; set; }
    public BlogPost Post { get; set; } = null!;

    // ========== LTREE PATH ==========

    // The hierarchical path in ltree format
    // Format: ancestor1.ancestor2.thisNode
    // Examples:
    //   Root comment: "1"
    //   Child of 1: "1.5"
    //   Grandchild: "1.5.12"
    //
    // Using the LTree type enables LINQ translations for ltree operators
    public LTree Path { get; set; }

    // Keep ParentCommentId for convenience
    public int? ParentCommentId { get; set; }
    public Comment? ParentComment { get; set; }
    public ICollection<Comment> Children { get; set; } = new List<Comment>();

    // ========== HELPER METHODS ==========

    // Helper to get depth - LTree has NLevel property for this
    public int GetDepth() => Path.NLevel - 1;

    public IEnumerable<int> GetAncestorIds()
    {
        var pathString = Path.ToString();
        if (string.IsNullOrEmpty(pathString)) yield break;

        var parts = pathString.Split('.');
        // All except last (which is this node)
        for (int i = 0; i < parts.Length - 1; i++)
        {
            if (int.TryParse(parts[i], out var id))
                yield return id;
        }
    }
}

EF Core Configuration

public class CommentConfiguration : IEntityTypeConfiguration<Comment>
{
    public void Configure(EntityTypeBuilder<Comment> builder)
    {
        builder.HasKey(c => c.Id);

        builder.Property(c => c.Content)
            .IsRequired()
            .HasMaxLength(10000);

        builder.Property(c => c.Author)
            .IsRequired()
            .HasMaxLength(200);

        // ========== PATH COLUMN ==========
        // The LTree type is automatically mapped to PostgreSQL's ltree type
        // by the Npgsql provider - no explicit column type needed
        builder.Property(c => c.Path)
            .IsRequired();

        // Relationship to blog post
        builder.HasOne(c => c.Post)
            .WithMany(p => p.Comments)
            .HasForeignKey(c => c.PostId)
            .OnDelete(DeleteBehavior.Cascade);

        // Self-referencing
        builder.HasOne(c => c.ParentComment)
            .WithMany(c => c.Children)
            .HasForeignKey(c => c.ParentCommentId)
            .OnDelete(DeleteBehavior.Restrict);

        // Standard indexes
        builder.HasIndex(c => c.PostId);
        builder.HasIndex(c => c.ParentCommentId);
    }
}

Add the GiST index via migration:

protected override void Up(MigrationBuilder migrationBuilder)
{
    // GiST index for ltree - enables efficient @>, <@, and ~ operators
    migrationBuilder.Sql(
        "CREATE INDEX ix_comments_path_gist ON comments USING GIST (path)");

    // Alternative: B-tree index for exact match and sorting
    // migrationBuilder.Sql(
    //     "CREATE INDEX ix_comments_path_btree ON comments USING BTREE (path)");
}

ltree Operators

ltree provides powerful operators. The Npgsql EF Core provider translates LTree methods to these operators:

Additional LINQ-translatable properties and methods:

• ltree.NLevel → nlevel(ltree) - number of labels in path
• ltree.Subtree(start, end) → subltree(ltree, start, end) - extract range of labels
• ltree.Subpath(offset) → subpath(ltree, offset) - suffix from offset
• ltree.Subpath(offset, len) → subpath(ltree, offset, len) - substring
• ltree.Index(subpath) → index(ltree, subpath) - find subpath position
• LTree.LongestCommonAncestor(ltree1, ltree2) → lca(ltree1, ltree2) - lowest common ancestor

Operations

Insert a New Comment

public async Task<Comment> AddCommentAsync(
    int postId,
    int? parentId,
    string author,
    string content,
    CancellationToken ct = default)
{
    string path;

    if (parentId.HasValue)
    {
        // Get parent's path
        var parentPath = await context.Comments
            .Where(c => c.Id == parentId.Value)
            .Select(c => c.Path)
            .FirstOrDefaultAsync(ct);

        if (parentPath == null)
            throw new InvalidOperationException($"Parent comment {parentId} not found");

        // Create comment first to get the ID
        var comment = new Comment
        {
            PostId = postId,
            ParentCommentId = parentId,
            Author = author,
            Content = content,
            CreatedAt = DateTime.UtcNow,
            Path = string.Empty  // Temporary
        };

        context.Comments.Add(comment);
        await context.SaveChangesAsync(ct);

        // Build path: parentPath.newId
        // ltree uses periods as separators
        comment.Path = $"{parentPath}.{comment.Id}";
        await context.SaveChangesAsync(ct);

        logger.LogInformation("Added comment {CommentId} with ltree path {Path}",
            comment.Id, comment.Path);
        return comment;
    }
    else
    {
        // Root comment - path is just the ID
        var comment = new Comment
        {
            PostId = postId,
            ParentCommentId = null,
            Author = author,
            Content = content,
            CreatedAt = DateTime.UtcNow,
            Path = string.Empty
        };

        context.Comments.Add(comment);
        await context.SaveChangesAsync(ct);

        comment.Path = comment.Id.ToString();
        await context.SaveChangesAsync(ct);

        return comment;
    }
}

Get Immediate Children

Using ParentCommentId (simple) or ltree pattern matching:

public async Task<List<Comment>> GetChildrenAsync(int commentId, CancellationToken ct = default)
{
    // Option 1: Simple ParentCommentId lookup
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.ParentCommentId == commentId)
        .OrderBy(c => c.CreatedAt)
        .ToListAsync(ct);
}

// Option 2: Using ltree pattern (demonstration)
public async Task<List<Comment>> GetChildrenLtreeAsync(int commentId, CancellationToken ct = default)
{
    // Get parent path first
    var parentPath = await context.Comments
        .Where(c => c.Id == commentId)
        .Select(c => c.Path)
        .FirstOrDefaultAsync(ct);

    if (parentPath == null)
        return new List<Comment>();

    // Children match pattern: parentPath.*{1}
    // The {1} means exactly one more label (immediate children only)
    var sql = @"
        SELECT * FROM comments
        WHERE path ~ ($1 || '.*{1}')::lquery
        ORDER BY created_at";

    return await context.Comments
        .FromSqlRaw(sql, parentPath)
        .AsNoTracking()
        .ToListAsync(ct);
}

Get All Ancestors

Using LINQ with the IsAncestorOf method (translates to @> operator):

public async Task<List<Comment>> GetAncestorsAsync(int commentId, CancellationToken ct = default)
{
    var targetPath = await context.Comments
        .Where(c => c.Id == commentId)
        .Select(c => c.Path)
        .FirstOrDefaultAsync(ct);

    if (targetPath == default)
        return new List<Comment>();

    // Find all nodes whose path is an ancestor of this path
    // Using IsAncestorOf which translates to @> operator
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.IsAncestorOf(targetPath) && c.Id != commentId)
        .OrderBy(c => c.Path.NLevel)
        .ToListAsync(ct);
}

Get All Descendants

Using LINQ with the IsDescendantOf method (translates to <@ operator):

public async Task<List<Comment>> GetDescendantsAsync(int commentId, CancellationToken ct = default)
{
    var parentPath = await context.Comments
        .Where(c => c.Id == commentId)
        .Select(c => c.Path)
        .FirstOrDefaultAsync(ct);

    if (parentPath == default)
        return new List<Comment>();

    // Find all nodes whose path is a descendant of this path
    // Using IsDescendantOf which translates to <@ operator
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.IsDescendantOf(parentPath) && c.Id != commentId)
        .OrderBy(c => c.Path)
        .ToListAsync(ct);
}

Get Descendants to Maximum Depth

Using LINQ with NLevel for depth limiting:

public async Task<List<Comment>> GetDescendantsToDepthAsync(
    int commentId,
    int maxDepth,
    CancellationToken ct = default)
{
    var comment = await context.Comments
        .FirstOrDefaultAsync(c => c.Id == commentId, ct);

    if (comment == null)
        return new List<Comment>();

    var basePath = comment.Path;
    var baseLevel = comment.Path.NLevel;

    // NLevel property translates to nlevel() function
    // Filter descendants within maxDepth levels
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.IsDescendantOf(basePath) 
                 && c.Id != commentId
                 && c.Path.NLevel - baseLevel <= maxDepth)
        .OrderBy(c => c.Path)
        .ToListAsync(ct);
}

// If you need the depth value in results, you can project it:
public async Task<List<CommentWithDepth>> GetDescendantsWithDepthAsync(
    int commentId,
    int maxDepth,
    CancellationToken ct = default)
{
    var comment = await context.Comments
        .FirstOrDefaultAsync(c => c.Id == commentId, ct);

    if (comment == null)
        return new List<CommentWithDepth>();

    var basePath = comment.Path;
    var baseLevel = comment.Path.NLevel;

    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.IsDescendantOf(basePath) 
                 && c.Id != commentId
                 && c.Path.NLevel - baseLevel <= maxDepth)
        .OrderBy(c => c.Path)
        .Select(c => new CommentWithDepth
        {
            Id = c.Id,
            Content = c.Content,
            Author = c.Author,
            CreatedAt = c.CreatedAt,
            PostId = c.PostId,
            ParentCommentId = c.ParentCommentId,
            Path = c.Path.ToString(),
            Depth = c.Path.NLevel - baseLevel
        })
        .ToListAsync(ct);
}

Pattern Matching Queries

ltree supports powerful lquery patterns. Use MatchesLQuery in LINQ:

// Find all comments at exactly depth 2 under comment 1
public async Task<List<Comment>> GetAtDepthAsync(int commentId, int depth, CancellationToken ct = default)
{
    var path = await context.Comments
        .Where(c => c.Id == commentId)
        .Select(c => c.Path)
        .FirstOrDefaultAsync(ct);

    if (path == default) return new List<Comment>();

    // Pattern: path.*{depth} matches exactly 'depth' more levels
    var pattern = $"{path}.*{{{depth}}}";
    
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.MatchesLQuery(pattern))
        .OrderBy(c => c.Path)
        .ToListAsync(ct);
}

// Find all paths matching a pattern like "1.*.7" (any path through 1 ending in 7)
public async Task<List<Comment>> MatchPatternAsync(string pattern, CancellationToken ct = default)
{
    // MatchesLQuery translates to the ~ operator
    return await context.Comments
        .AsNoTracking()
        .Where(c => c.Path.MatchesLQuery(pattern))
        .OrderBy(c => c.Path)
        .ToListAsync(ct);
}

Delete a Subtree

You can use LINQ to select the subtree, then delete:

public async Task DeleteSubtreeAsync(int commentId, CancellationToken ct = default)
{
    var path = await context.Comments
        .Where(c => c.Id == commentId)
        .Select(c => c.Path)
        .FirstOrDefaultAsync(ct);

    if (path == default)
        throw new InvalidOperationException($"Comment {commentId} not found");

    // Delete all descendants (nodes where path is descendant of this path)
    // Note: ExecuteDeleteAsync requires EF Core 7+
    var deleted = await context.Comments
        .Where(c => c.Path.IsDescendantOf(path))
        .ExecuteDeleteAsync(ct);

    logger.LogInformation("Deleted {Count} comments with path prefix {Path}", deleted, path);
}

Move a Subtree

ltree provides functions to help with path manipulation:

public async Task MoveSubtreeAsync(
    int commentId,
    int newParentId,
    CancellationToken ct = default)
{
    await using var transaction = await context.Database.BeginTransactionAsync(ct);

    try
    {
        var node = await context.Comments.FirstOrDefaultAsync(c => c.Id == commentId, ct);
        var newParent = await context.Comments.FirstOrDefaultAsync(c => c.Id == newParentId, ct);

        if (node == null || newParent == null)
            throw new InvalidOperationException("Node or parent not found");

        // Prevent cycles
        if (newParent.Path.StartsWith(node.Path))
            throw new InvalidOperationException("Cannot move under own descendant");

        var oldPath = node.Path;
        var newPath = $"{newParent.Path}.{node.Id}";

        // Update all descendants: replace old path prefix with new one
        // subpath(path, nlevel(oldPath)) gets the suffix after oldPath
        // We concatenate newPath with that suffix
        var sql = @"
            UPDATE comments
            SET path = $2::ltree || subpath(path, nlevel($1::ltree))
            WHERE path <@ $1::ltree";

        await context.Database.ExecuteSqlRawAsync(
            sql,
            new object[] { oldPath, newPath },
            ct);

        // Update parent reference
        node.ParentCommentId = newParentId;
        await context.SaveChangesAsync(ct);

        await transaction.CommitAsync(ct);

        logger.LogInformation("Moved subtree from {OldPath} to {NewPath}", oldPath, newPath);
    }
    catch
    {
        await transaction.RollbackAsync(ct);
        throw;
    }
}

ltree Functions Reference

PostgreSQL provides many useful ltree functions:

Query Flow Visualisation

sequenceDiagram
    participant App as Application
    participant EF as EF Core
    participant PG as PostgreSQL + ltree

    Note over App,PG: Getting Descendants (GiST index)
    App->>EF: GetDescendantsAsync(commentId)
    EF->>PG: SELECT path FROM comments WHERE id = @id
    PG-->>EF: Path "1.3"
    EF->>PG: SELECT * FROM comments WHERE path <@ '1.3'::ltree
    Note over PG: Uses GiST index - O(log n)
    PG-->>EF: All descendants
    EF-->>App: List<Comment>

    Note over App,PG: Pattern Match Query
    App->>EF: MatchPatternAsync("1.*.7")
    EF->>PG: SELECT * FROM comments WHERE path ~ '1.*.7'::lquery
    Note over PG: GiST index supports pattern matching
    PG-->>EF: Matching comments
    EF-->>App: List<Comment>

Performance Characteristics

With GiST indexes, ltree queries are extremely efficient - typically O(log n) regardless of tree depth.

Pros and Cons

When to Use ltree

Choose ltree when:

• You're committed to PostgreSQL
• Performance is critical for hierarchy queries
• You need pattern matching (find all X.*.Y paths)
• You want the best of materialised paths
• You want LINQ support for most hierarchy operations

Avoid ltree when:

• You need database portability (SQL Server, MySQL, etc.)
• Your team is unfamiliar with PostgreSQL extensions
• Labels need non-alphanumeric characters
• You need recursive CTEs and want to avoid any raw SQL

Comparison with Materialised Path

Example: Full Comment Tree Query

Putting it all together - get an entire comment tree with depth for a blog post:

public async Task<List<CommentTreeItem>> GetPostCommentTreeAsync(
    int postId,
    int maxDepth = 5,
    CancellationToken ct = default)
{
    // Get all comments for the post with calculated depth
    // nlevel() counts the labels in the path
    var sql = @"
        WITH root_comments AS (
            -- Find root comments for this post (no dot in path = root)
            SELECT path, nlevel(path) as root_level
            FROM comments
            WHERE post_id = $1 AND path !~ '*.*'
        )
        SELECT
            c.id,
            c.content,
            c.author,
            c.created_at,
            c.post_id,
            c.parent_comment_id,
            c.path::text as path,
            nlevel(c.path) - COALESCE(
                (SELECT root_level FROM root_comments r
                 WHERE c.path <@ r.path
                 ORDER BY nlevel(r.path) DESC LIMIT 1),
                nlevel(c.path)
            ) as depth
        FROM comments c
        WHERE c.post_id = $1
          AND nlevel(c.path) <= $2 + 1  -- +1 because depth is 0-indexed
        ORDER BY c.path";  -- Perfect depth-first order!

    return await context.Database
        .SqlQueryRaw<CommentTreeItem>(sql, postId, maxDepth)
        .ToListAsync(ct);
}

public class CommentTreeItem
{
    public int Id { get; set; }
    public string Content { get; set; } = string.Empty;
    public string Author { get; set; } = string.Empty;
    public DateTime CreatedAt { get; set; }
    public int PostId { get; set; }
    public int? ParentCommentId { get; set; }
    public string Path { get; set; } = string.Empty;
    public int Depth { get; set; }
}

Series Navigation

• Part 1: Overview
• Part 1.1: Adjacency List
• Part 1.2: Closure Table
• Part 1.3: Materialised Path
• Part 1.4: Nested Sets
• Part 1.5: ltree (this article)

What's Next?

This series has covered five approaches to hierarchical data using EF Core. Part 2 will explore using raw SQL and Dapper for even more control over hierarchy queries - coming soon!