admin/PowerShell-PowerShell
1
0
Fork 0
mirror of https://github.com/PowerShell/PowerShell synced 2026-06-08 12:12:50 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update coding guidelines (#4754)

Browse Source 
* Update coding guidelines to make it more concrete and useful in a review process.

* Add a section about security

* Address comments

* Update .spelling

* Address more comments

* Address comment about the security section

* Fix 'i.e.' and 'e.g.'

* Update .spelling

* One more tweak

* Add motivation for using 'nameof'
This commit is contained in:
Dongbo Wang
2017-09-08 20:24:26 +04:00
committed by Ilya
parent fc9d798b77
commit ca1d2acfba
4 changed files with 200 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/dev-process/coding-guidelines.md
+180 -68
View File
@@ -1,92 +1,204 @@
# C# Coding Style
# C# Coding Guidelines
## Coding Conventions
As a general rule, our coding convention is to follow the style of the surrounding code.
Avoid reformatting any code when submitting a PR as it obscures the functional changes of your change.
We run the [.NET code formatter tool](https://github.com/dotnet/codeformatter) regularly help keep consistent formatting.
So if a file happens to differ in style from conventions defined here
(e.g. private members are named `m_member` rather than `_member`),
the existing style in that file takes precedence.
A basic rule of formatting is to use "Visual Studio defaults".
Here are some general guidelines
When making changes, you may find some existing code goes against the conventions defined here.
In such cases, please avoid reformatting any existing code when submitting a PR as it obscures the functional changes of the PR.
A separate PR should be submitted for style-only changes.
We also run the [.NET code formatter tool](https://github.com/dotnet/codeformatter) regularly to keep consistent formatting.
* No tabs, indent 4 spaces.
* Braces usually go on their own line,
### Naming Conventions
* Use meaningful, descriptive words for names.
For method names, it's encouraged to use `VerbObject` pair such as **`LoadModule`**.
* Use `_camelCase` to name internal and private fields and use `readonly` where possible.
Prefix instance fields with `_`, static fields with `s_` and thread static fields with `t_`.
When used on static fields, `readonly` should come after `static` (i.e. `static readonly` not `readonly static`).
* Use `camelCase` to name non-constant local variables.
* Use `PascalCase` to name constant local variables and fields.
The only exception is for interop code where the constant should exactly match the name and value of the code you are calling via interop (e.g. `const int ERROR_SUCCESS = 0`).
* Use `PascalCase` to name types and all other type members.
### Layout Conventions
* Use four spaces of indentation (no tabs).
* Avoid more than one blank empty line at any time.
* Avoid trailing spaces at the end of a line.
* Braces usually go on their own lines,
with the exception of single line statements that are properly indented.
* Use `_camelCase` for instance fields,
use `readonly` where possible.
* Namespace imports should be specified at the top of the file,
outside of `namespace` declarations.
* Fields should be specified at the top within type declarations.
For those that serve as backing fields for properties,
they should be specified next to the corresponding properties.
* Preprocessor directives like `#if` and `#endif` should be placed at the beginning of a line,
without any leading spaces.
* File encoding should be `ASCII`.
All `BOM` encodings should be avoided.
Tests that need a `BOM` encoding file should generate the file on the fly.
### Member Conventions
* Use of `this` is neither encouraged nor discouraged.
* Avoid more than one blank empty line.
* Public members should use [doc comments](https://msdn.microsoft.com/en-us/library/b2s063f7.aspx),
internal members may use doc comments but it is not encouraged.
* Use `nameof(<member-name>)` instead of `"<member-name>"` whenever possible and relevant.
The motivation is to easily and more accurately find references.
* Always specify the visibility, even if it's the default (i.e. `private string _foo` not `string _foo`).
Visibility should be the first modifier (i.e. `public abstract` not `abstract public`).
* Make members private where possible.
Avoid declaring public members unless it's absolutely necessary.
* Public members in a namespace that ends with `Internal`,
for example `System.Management.Automation.Internal` are not considered a supported public API.
Such members are necessarily public as implementation details in code shared between C# and PowerShell script,
or must be available publicly by generated code.
* File encoding should be ASCII (preferred)
or UTF8 (with BOM) if absolutely necessary.
## Preprocessor defines
### Commenting Conventions
There are 3 primary preprocessor macros we define during builds:
* Place the comment on a separate line, not at the end of a line of code.
* DEBUG - guard code that should not be included in release builds
* CORECLR - guard code that differs between Full CLR and CoreCLR
* UNIX - guard code that is specific to Unix (Linux and macOS)
* Begin comment text with an uppercase letter.
It's recommended to end comment text with a period but not required.
* Add comments where the code is not trivial or could be confusing.
* Add comments where a reviewer needs help to understand the code.
* Update/remove existing comments when you are changing the corresponding code.
* Make sure the added/updated comments are meaningful, accurate and easy to understand.
* Public members must use [doc comments](https://msdn.microsoft.com/en-us/library/b2s063f7.aspx).
Internal and private members may use doc comments but it is not required.
## Performance Considerations
PowerShell has a lot of performance sensitive code as well as a lot of inefficient code.
We have some guidelines that we typically apply widely even in less important code because code and patterns are copied,
and we want certain inefficient code to stay out of the performance critical code.
Some general guidelines:
* Avoid LINQ - it can create lots of avoidable garbage.
Instead, iterate through a collection directly using `for` or `foreach` loop.
* Between `for` and `foreach`,
`for` is slightly preferred when you're uncertain if `foreach` allocates an iterator.
* Avoid `params` arrays, prefer adding overloads with 1, 2, 3, and maybe more parameters.
* Be aware of APIs such as `String.Split(params char[])` that do not provide overloads to avoid array allocation.
When calling such APIs, reuse a static array when possible (e.g. `Utils.Separators.Colon`).
* Avoid creating empty arrays.
Instead, reuse the static ones via `Utils.EmptyArray<T>`.
* Avoid unnecessary memory allocation in a loop.
Move the memory allocation outside the loop if possible.
* Avoid gratuitous exceptions as much as possible.
Exception handling can be expensive due to cache misses and page faults when accessing the handling code and data.
Finding and designing away exception-heavy code can result in a decent performance win.
For example, you should stay away from things like using exceptions for control flow.
* Avoid `if (obj is Example) { example = (Example)obj }` when casting an object to a type.
Instead, use `var example = obj as Example` or the C# 7 syntax `if (obj is Example example) {...}` as appropriate.
In this way you can avoid converting to the type twice.
* Use `dict.TryGetValue` instead of `dict.Contains` and `dict[..]` when retrieving value from a `Dictionary`.
In this way you can avoid hashing the key twice.
* It's OK to use the `+` operator to concatenate one-off short strings.
But when dealing with strings in loops or large amounts of text,
use a `StringBuilder` object.
## Security Considerations
Security is an important aspect of PowerShell and we need to be very careful about changes that may introduce security risks,
such as code injection caused by the lack of input validation,
privilege escalation due to the misuse of impersonation,
or data privacy breach with a plain text password.
Reviewers of a PR should be sensitive to changes that may affect security.
Some security related keywords may serve as good indicators,
such as `password`, `crypto`, `encryption`, `decryption`, `certificate`, `authenticate`, `ssl/tls` and `protected data`.
When facing a PR with such changes,
the reviewers should request a designated security Subject Matter Expert (SME) to review the PR.
Currently, @PaulHigin and @TravisEz13 are our security SMEs.
See [CODEOWNERS](../../.github/CODEOWNERS) for more information about the area experts.
## Best Practices
* Avoid hard-coding anything unless it's absolutely necessary.
* Avoid a method that is too long and complex.
In such case, separate it to multiple methods or even a nested class as you see fit.
* Use `using` statement instead of `try/finally` if the only code in the `finally` block is to call the `Dispose` method.
* Use of object initializers (e.g. `new Example { Name = "Name", ID = 1 }`) is encouraged for better readability,
but not required.
* Stick to the `DRY` principle -- Don't Repeat Yourself.
* Wrap the commonly used code in methods,
or even put it in a utility class if that makes sense,
so that the same code can be reused (e.g. `StringToBase64Converter.Base64ToString(string)`).
* Check if the code for the same purpose already exists in the code base before inventing your own wheel.
* Avoid repeating literal strings in code. Instead, use `const` variable to hold the string.
* Resource strings used for errors or UI should be put in resource files (`.resx`) so that they can be localized later.
* Use of new C# language syntax is encouraged.
But avoid refactoring any existing code using new language syntax when submitting a PR
as it obscures the functional changes of the PR.
A separate PR should be submitted for such refactoring without any functional changes.
## Portable Code
There are 3 primary preprocessor macros we use during builds:
* `DEBUG` - guard code that should not be included in release builds
* `CORECLR` - guard code that differs between Full CLR and CoreCLR
* `UNIX` - guard code that is specific to Unix (Linux and macOS)
Any other preprocessor defines found in the source are used for one-off custom builds,
typically to help debug specific scenarios.
### Runtimes
Here are some general guidelines for writing portable code:
The PowerShell repo is used to build PowerShell targeting CoreCLR as well as CLR 4.5.
* We are in the process of cleaning up Full CLR specific code (code enclosed in `!CORECLR`),
so do not use `CORECLR` or `!CORECLR` in new code.
PowerShell Core targets .NET Core only and all new changes should support .NET Core only.
Code under !CORECLR must build against CLR 4.5.
We will not accept changes that require a later version of the full CLR.
In extremely rare cases, we may use reflection to use an API in a later version of the CLR,
but the feature must robustly handle running with CLR 4.5.
* The PowerShell code base started on Windows and depends on many Win32 APIs through P/Invoke.
Going forward, we try to depend on .NET Core to handle platform differences,
so avoid adding new P/Invoke calls where a suitable alternative exists in .NET Core.
We may reject code under !CORECLR without explanation because
we do not support installation or testing of such code in this repo.
All new features should support CoreCLR.
* Try to minimize the use of `#if UNIX`.
When absolutely necessary, avoid duplicating more code than necessary,
and instead prefer introducing helper functions to minimize the platform differences.
## Performance considerations
PowerShell has a lot of performance sensitive code as well as a lot of inefficient code.
We have some guidelines that we typically apply widely even in less important code
because code and patterns are copied we want certain inefficient code to stay out of the performance critical code.
Some general guidelines:
* Avoid LINQ - it can create lots of avoidable garbage
* Prefer `for` and `foreach`,
with a slight preference towards `for` when you're uncertain if `foreach` allocates an iterator.
* Avoid `params` arrays, prefer adding overloads with 1, 2, 3, and maybe more parameters.
* Be aware of APIs such as `String.Split(params char[])` that do not provide overloads to avoid array allocation.
When calling such APIs, reuse a static array when possible.
* Avoid unnecessary memory allocation in a loop.
Move the memory allocation outside the loop if possible.
## Portable code
The PowerShell code base started on Windows and depends on many Win32 APIs through P/Invoke.
Going forward, we try to depend on CoreCLR to handle platform differences,
so avoid adding new P/Invoke calls where a suitable alternative exists in .NET.
Try to minimize the use of `#if UNIX` and `#if CORECLR`.
When absolutely necessary, avoid duplicating more code than necessary,
and instead prefer introducing helper functions to minimize the platform differences.
When adding platform dependent code, prefer preprocessor directives
over runtime checks.
We produce a single binary for all UNIX variants,
so runtime checks are currently necessary for some platform differences, e.g. macOS and Linux.
## Code comments
It's strongly encouraged to add comments when you are making changes to the code and tests,
especially when the changes are not trivial or may raise confusion.
Make sure the added comments are accurate and easy to understand.
Good code comments would greatly improve readability of the code, and make it much more maintainable.
* When adding platform dependent code (`Windows` vs. `UNIX`), prefer preprocessor directives over runtime checks.
However, runtime checks are acceptable if it would greatly improve readability
without causing performance concerns in performance-sensitive code.
* We produce a single binary for all UNIX variants,
so runtime checks are currently necessary for some of them (e.g. macOS vs. Linux).