Which type should I use in C# to represent numbers? - Luca Bolognese

Which type should I use in C# to represent numbers?

Luca -

☕ 1 min. read

Yesterday I found an old email in my mail box that I thought might be gen­er­ally in­ter­est­ing.

I was ask­ing the tech­ni­cal lead on the C# com­piler which al­go­rithm/​short­cut peo­ple should use to choose their ‘number types’ among the many avail­able in the lan­guage. I was ask­ing for some­thing that works the ma­jor­ity of times, even if not al­ways. I’m sure there are other sce­nar­ios we haven’t con­sider. Anyhow, here is his al­go­rithm.

If you need frac­tions:

  • Use dec­i­mal when in­ter­me­di­ate re­sults need to be rounded to fixed pre­ci­sion - this is al­most al­ways lim­ited to cal­cu­la­tions in­volv­ing money.
  • Otherwise use dou­ble - you will get the round­ing of your cal­cu­la­tions wrong, but the ex­tra pre­ci­sion of dou­ble will en­sure that your re­sults will be good enough.
  • Only use float if you know you have a space is­sue, and you know the pre­ci­sion im­pli­ca­tions. If you don’t have a PhD in nu­meric com­pu­ta­tion you don’t qual­ify.

Otherwise:

  • Use int when­ever your val­ues can fit in an int, even for val­ues which can never be neg­a­tive. This is so that sub­trac­tion op­er­a­tions don’t get you con­fused.
  • Use long when your val­ues can’t fit in an int.

Byte, sbyte, short, ushort, uint, and ulong should only ever be used for in­terop with C code. Otherwise they’re not worth the has­sle.

12 Comments

Comments

Cory Nelson

2007-02-27T18:29:35Z

i really mean this in the nicest way: if you get confused by unsigned types, stop coding.  maybe the technical lead needs to have more faith in the competency of the developers he is targeting.

Sean

2007-02-27T19:12:35Z

Yeah, but in the .NET framework unsigned types are not CLR compliant, which you might or might not be worried about.

Fabrice's weblog

2007-02-27T19:24:05Z

Luca Bolognese, from the Microsoft C# team, has an interesting post that aims at providing answers to

NBC

2007-02-27T20:35:24Z

but if you use what the value means, isn't it more  readable ?
I mean if you have a file header
with a uint32 fourcc then a uint16 version value, its more readable since you know what you should look for. if you use int fourcc and int version i don"t know how the bytes map in the file.

Eber Irigoyen

2007-02-27T20:59:12Z

but this applies to the whole CTS, not just C#... right?

.NET a 2.860 m de altura

2007-02-28T22:57:25Z

C# nos ofrece varios tipos de datos para representar números y, sobretodo para quienes están

There Must Be Some Mistake

2007-03-01T20:54:07Z

Saw this post today describing basic guidance for when to use the different numeric data types available

Charlie Calvert's Community Bl

2007-03-02T03:23:24Z

Welcome to the twenty-second Community Convergence, the March CTP issue. I'm Charlie Calvert, the C#

RSS It All

2007-03-02T03:37:00Z

Welcome to the twenty-second Community Convergence, the March CTP issue. I&#039m Charlie Calvert, the

Ronaldo Ferreira

2007-03-02T18:36:41Z

I coded these days a method that reads a binary file, but all data were represented with only two bytes (an ushort is enough). I really do not see the reason to "forget" these types. If I have coded this same method with int, I have had throw away a lot of bytes.

NBC

2007-03-03T10:34:00Z

>I have had throw away a lot of bytes.
internaly it maybe have been casted as int anyway by the compilator

AS

2007-03-11T07:34:06Z

I want to write a Matrix (Math) with generics that can instance it with all numbers(float, double, int,...).
any idea?
I think it is imposible.
[imposible says i am posible :) ]

0 Webmentions

These are webmentions via the IndieWeb and webmention.io.