Pandatech.EFCore.PostgresExtensions

PostgreSQL-specific extensions for Entity Framework Core that fill the gaps left by the official Npgsql provider.

Feature	What it does
Row-level locking	FOR UPDATE with Wait, SkipLocked, and NoWait behaviors
Random-increment IDs	Non-predictable sequential IDs backed by a PostgreSQL sequence
Natural sort keys	Human-friendly ordering for strings containing numbers
Schema rollback helpers	Clean Down() migration methods for all of the above

Targets net8.0, net9.0, and net10.0.

Installation

dotnet add package Pandatech.EFCore.PostgresExtensions

Row-level locking

PostgreSQL's FOR UPDATE clause lets you lock selected rows for the duration of a transaction. This package exposes it as a LINQ extension method with three lock behaviors.

Setup

Register the query interceptor on your DbContext:

services.AddDbContextPool<AppDbContext>(options => options.UseNpgsql(connectionString) .UseQueryLocks());

Usage

The ForUpdate method must be called inside a transaction:

await using var transaction = await dbContext.Database.BeginTransactionAsync(); var order = await dbContext.Orders .Where(o => o.Id == orderId) .ForUpdate(LockBehavior.SkipLocked) .FirstOrDefaultAsync(); // Modify the locked row order.Status = OrderStatus.Processing; await dbContext.SaveChangesAsync(); await transaction.CommitAsync();

Lock behaviors

Behavior	SQL generated	When to use
Default (Wait)	FOR UPDATE	You need the row and can wait for it
SkipLocked	FOR UPDATE SKIP LOCKED	Queue-style processing — skip rows another worker already holds
NoWait	FOR UPDATE NOWAIT	Fail immediately if the row is locked

Random-increment sequence IDs

Generates bigint IDs that increment by a random amount within a configurable range. The IDs are unique and always increasing, but the gaps between them are unpredictable — preventing enumeration attacks while keeping an index-friendly insert order.

1. Configure the entity

public class Animal { public long Id { get; set; } public string Name { get; set; } } public class AnimalConfiguration : IEntityTypeConfiguration<Animal> { public void Configure(EntityTypeBuilder<Animal> builder) { builder.HasKey(x => x.Id); builder.Property(x => x.Id) .HasRandomIdSequence(); } }

2. Create the function in a migration

The function must be created before the table that references it. Add it manually at the top of your Up() method:

protected override void Up(MigrationBuilder migrationBuilder) { // Create the sequence + function first migrationBuilder.CreateRandomIdSequence( tableName: "animal", pkName: "id", startValue: 5, minRandIncrementValue: 5, maxRandIncrementValue: 10); // Then create the table (the default value references the function) migrationBuilder.CreateTable(...); } protected override void Down(MigrationBuilder migrationBuilder) { migrationBuilder.DropRandomIdSequence("animal", "id"); migrationBuilder.DropTable("animal"); }

Parameters: startValue is the first ID returned, minRandIncrementValue and maxRandIncrementValue define the random gap range between consecutive IDs.

Natural sort keys

Sorts strings that contain numbers the way a human would expect: "Item 2" before "Item 10", not after it. The package creates a PostgreSQL function that zero-pads numeric substrings, producing a sortable text key.

1. Create the function (once per database)

protected override void Up(MigrationBuilder migrationBuilder) { migrationBuilder.CreateNaturalSortKeyFunction(); } protected override void Down(MigrationBuilder migrationBuilder) { migrationBuilder.DropNaturalSortKeyFunction(); }

2. Add a computed column

public class Building { public long Id { get; set; } public string Address { get; set; } public string AddressNaturalSortKey { get; set; } } public class BuildingConfiguration : IEntityTypeConfiguration<Building> { public void Configure(EntityTypeBuilder<Building> builder) { builder.Property(x => x.AddressNaturalSortKey) .HasNaturalSortKey("address"); } }

Then order by the computed column:

var sorted = await dbContext.Buildings .OrderBy(b => b.AddressNaturalSortKey) .ToListAsync();

Encrypted-column unique indexes

For columns that store encrypted data where only the first 64 characters are deterministic (e.g., hash prefix), you can create a unique index on that prefix:

// In migration Up() migrationBuilder.CreateUniqueIndexOnEncryptedColumn("users", "email_encrypted"); // With an optional WHERE condition migrationBuilder.CreateUniqueIndexOnEncryptedColumn("users", "email_encrypted", "is_active = true"); // In migration Down() migrationBuilder.DropUniqueIndexOnEncryptedColumn("users", "email_encrypted");

API reference

Extension methods

Method	Target	Description
UseQueryLocks()	DbContextOptionsBuilder	Registers the interceptor that rewrites tagged queries into FOR UPDATE SQL
ForUpdate()	IQueryable<T>	Tags a query for row-level locking (must be inside a transaction)
HasRandomIdSequence()	PropertyBuilder	Configures the property to use a random-increment sequence as its default value
HasNaturalSortKey(column)	PropertyBuilder	Configures the property as a stored computed column using the natural sort function
CreateRandomIdSequence(...)	MigrationBuilder	Creates the PostgreSQL sequence and generator function
DropRandomIdSequence(...)	MigrationBuilder	Drops the sequence and generator function
CreateNaturalSortKeyFunction()	MigrationBuilder	Creates the natural sort key function
DropNaturalSortKeyFunction()	MigrationBuilder	Drops the natural sort key function
CreateUniqueIndexOnEncryptedColumn(...)	MigrationBuilder	Creates a unique index on the first 64 chars of a column
DropUniqueIndexOnEncryptedColumn(...)	MigrationBuilder	Drops the encrypted-column unique index

Enums

LockBehavior: Default (wait), SkipLocked, NoWait.

License

MIT