Multithreaded Generation not working as expected

[ahmedwael216]

Bogus NuGet Package

35.6.5

.NET Version

.NET 9

Visual Studio Version

rider 2025.3

What operating system are you using?

Windows

What locale are you using with Bogus?

default,

Problem Description

I was facing some issues regarding slow generation with a relatively big faker that had a lot of nested entities.

so I tried parallel generation using parallel for each and using parallel enumerable, but both seemed to not change any thing. after finding this issue I tried changing my code to override the PopulateInternal method but found no difference at all

here are my benchmarks results(using benchmark dotnet).

Method	batchSize	Mean
'Normal Generation'	100	7.926 s
'Parallel Generation'	100	8.032 s
'Normal Generation'	1000	79.178 s
'Parallel Generation'	1000	79.887 s

Generally both methods scale with O(n) performance with the generation count.

I discovered this when trying to generate some data for benchmarks, that's why the count was large, for tests faker was fast enough. but for benchmarks generating the data takes ore than doing all the DB operations and it really struggled for several minutes when generating more than 10K

Note

My methods like AddressFaker.Create() are just static methods that return fakers for every entity with the appropriate rules

LINQPad Example or Reproduction Steps

	[Benchmark(Description = "Normal Generation")]
	[ArgumentsSource(nameof(BatchSizes))]
	public void Bench1(int batchSize)
	{
		var addressFaker = AddressFaker.Create();
		var buyerFaker = CompanyFaker.Create()
			.RuleFor(c => c.Addresses, (_, c) => addressFaker.RuleFor(a => a.CompanyId, c.EdcID).Generate(2));
		var orderEventFaker = EventFaker.CreateOrderFaker();
		var orderItemEventFaker = EventFaker.CreateOrderItemFaker();
		var customFieldOrderFaker = CustomFieldFaker.CreateOrderFaker();
		var customFieldOrderItemFaker = CustomFieldFaker.CreateOrderItemFaker();

		var orderItemBaseFaker = OrderItemFaker.Create();

		var orderFaker = OrderFaker.Create()
			.RuleFor(o => o.OwnerId, _ => OwnerId)
			.RuleFor(o => o.Buyer, _ => buyerFaker.Generate())
			.RuleFor(o => o.BuyerID, (_, o) => o.Buyer!.ID)
			.FinishWith((_, o) =>
			{
				o.Events = orderEventFaker.Generate(10);
				o.Events.ForEach(cf => cf.OrderId = o.ID);
				o.PackLines.ForEach(cf => cf.OrderId = o.ID);
				o.CustomFields = customFieldOrderFaker.Generate(10);
				o.CustomFields.ForEach(cf => cf.OrderId = o.ID);

				var items = orderItemBaseFaker.Generate(2);

				foreach (var item in items)
				{
					item.OrderId = o.ID;
					item.OwnerId = OwnerId;
					item.Events = orderItemEventFaker.Generate(10);
					item.Events.ForEach(cf => cf.OrderItemId = item.ID);
					item.CustomFields = customFieldOrderItemFaker.Generate(10);
					item.CustomFields.ForEach(cf => cf.OrderItemId = item.ID);
				}

				o.OrderLines = items;
			});

		var orders = orderFaker.Generate(batchSize);

		Console.WriteLine(orders.First().ID);
		Console.WriteLine(orders.Last().ID);
	}

	[Benchmark(Description = "Parallel Generation")]
	[ArgumentsSource(nameof(BatchSizes))]
	public void Bench2(int batchSize)
	{
		const int threadCount = 5;
		var batchSizePerThread = batchSize / threadCount;

		var addressFaker = AddressFaker.Create();
		var buyerFaker = CompanyFaker.Create()
			.RuleFor(c => c.Addresses, (_, c) => addressFaker.RuleFor(a => a.CompanyId, c.ID).Generate(2));

		var orderEventFaker = EventFaker.CreateOrderFaker();
		var orderItemEventFaker = EventFaker.CreateOrderItemFaker();
		var customFieldOrderFaker = CustomFieldFaker.CreateOrderFaker();
		var customFieldOrderItemFaker = CustomFieldFaker.CreateOrderItemFaker();

		var orderItemFaker = OrderItemFaker.Create()
			.RuleFor(it => it.Events, _ => orderItemEventFaker.Generate(10))
			.RuleFor(it => it.CustomFields, _ => customFieldOrderItemFaker.Generate(10));

		var baseOrderFaker = OrderFaker.Create()
			.RuleFor(o => o.OwnerId, _ => OwnerId)
			.RuleFor(o => o.Buyer, _ => buyerFaker.Generate())
			.RuleFor(o => o.BuyerID, (_, o) => o.Buyer!.ID)
			.RuleFor(o => o.Events, _ => orderEventFaker.Generate(10))
			.RuleFor(o => o.CustomFields, _ => customFieldOrderFaker.Generate(10))
			.RuleFor(o => o.OrderLines, _ => orderItemFaker.Generate(2))
			.FinishWith((_, o) =>
			{
				foreach (var ev in o.Events) ev.OrderId = o.ID;
				foreach (var pl in o.PackLines) pl.OrderId = o.ID;
				foreach (var cf in o.CustomFields) cf.OrderId = o.ID;

				foreach (var item in o.OrderLines)
				{
					item.OrderId = o.ID;
					item.OwnerId = OwnerId;

					foreach (var ev in item.Events) ev.OrderItemId = item.ID;
					foreach (var cf in item.CustomFields) cf.OrderItemId = item.ID;
				}
			});

		var orders = ParallelEnumerable.Range(0, threadCount)
			.SelectMany(_ =>baseOrderFaker.Clone().Generate(batchSizePerThread)).ToList();

		Console.WriteLine(orders.First().ID);
		Console.WriteLine(orders.Last().ID);
	}

Expected Behavior

parallel generation should be faster

Actual Behavior

both methods perform the same and scale the same

Known Workarounds

No response

Could you help with a pull-request?

Yes

[ahmedwael216]

Also as a note I am very open to help fix this issue if any of the maintainers can guide me what to fix

[ahmedwael216]

Update:
translated all my fakers to static functions that user the static functions and the numbers speak for themselves

Method	batchSize	Mean
'Normal Generation'	100	8,993.38 ms
'Static Generation'	100	60.44 ms
'Parallel Generation'	100	8,799.12 ms
'Normal Generation'	1000	83,236.50 ms
'Static Generation'	1000	528.29 ms
'Parallel Generation'	1000	86,359.68 ms

the only problem is this is good for benchmarks where I need a lot of data at once, but can't be used for tests as we need the flexibility if fluent methods to tailor the faker for the tests need

---

Update 2

seems trying to multithread static methods is not working at all and we are getting worse performance

Method	batchSize	Mean
'Static Generation'	1000	536.3 ms
'Parallel Generation'	1000	1,008.5 ms
'Static Generation'	10000	5,336.7 ms
'Parallel Generation'	10000	9,410.1 ms

notice that I generated 10x more entities here than previous benchmarks