Scott Hanselman

NuGet Package of the Week: Canopy Web Testing Framework with F#

March 25, '14 Comments [18] Posted in NuGet | NuGetPOW | Open Source
Sponsored By

I've been exploring Automated Browser Testing recently, and also checking out F# for unrelated reasons. However, when you combine the two you end up with "canopy." Canopy is a "f#rictionless web testing" framework that combines the flexibility of Selenium with the clean look of the F# language. F# is much terser (more elegant, even) than C#, and is garnering the interest of a lot of the .NET Open Source community. Folks are creating cool domain specific languages of their own using F# as the base.

You already have F# and perhaps didn't realize you did! If you don't, there's lots of ways to get F# for free. You can use F# for free with VS2013 Desktop Express plus Visual F# Tools 3.1.1.

F# is open source and cross platform, running on Linux, Mac OS X, Android, iOS, Windows as well as HTML5 and GPUs. F# is free to use and has an OSI-approved open source license.

Even if you don't feel like installing anything, you can learn and play with F# in your browser now! Check out http://www.tryfsharp.org 

Also check out FunScript, which is F# to JavaScript! Don't believe them? Try Pacman using F# and JavaScript with source!

image

Anyway, back to Canopy. Make a new Console app and NuGet in the canopy package:

image

The NuGet package will bring in Selenium as a dependency.

Then, try out their "Hello World" web testing sample, that I've also pasted here.

//these are similar to C# using statements
open canopy
open runner
open System

//start an instance of the firefox browser
start firefox

//this is how you define a test
"taking canopy for a spin" &&& fun _ ->
//this is an F# function body, it's whitespace enforced

//go to url
url "http://lefthandedgoat.github.io/canopy/testpages/"

//assert that the element with an id of 'welcome' has
//the text 'Welcome'
"#welcome" == "Welcome"

//assert that the element with an id of 'firstName' has the value 'John'
"#firstName" == "John"

//change the value of element with
//an id of 'firstName' to 'Something Else'
"#firstName" << "Something Else"

//verify another element's value, click a button,
//verify the element is updated
"#button_clicked" == "button not clicked"
click "#button"
"#button_clicked" == "button clicked"

//run all tests
run()

System.Console.WriteLine("press [enter] to exit")
System.Console.ReadLine() |> ignore

quit()

And boom, it just works. You can run this .NET application just like any other. .NET apps are .NET apps, as they say. It doesn't matter what language it's written in. When (if) you distribute this application you'd just include the contents of your Debug folder. No need to "install" F# or anything on the target machine.

image

You can do all sorts of Selenium testing with canopy, like:

//start a bunch of browsers and switch around
start firefox
let mainBrowser = browser
start chrome
let secondBrowser = browser
//switch back to mainBrowser after opening secondBrowser
switchTo mainBrowser

//take screenshots
let path = Environment.GetFolderPath(Environment.SpecialFolder.ApplicationData) + @"\canopy\"
let filename = DateTime.Now.ToString("MMM-d_HH-mm-ss-fff")
screenshot path filename

//get an element
element "#firstName" |> someParent

//press buttons
press tab
press enter
press down
press up
press left
press right

//check and click things
check "#yes"
click "#login"

//or even drag things!
drag ".todo" ".inprogress"

Oh, and by the way, the canopy library builds itself using FAKE, the F# Build System we talked about last week! Go check these projects out and offer to help or support them. There's a lot of interesting open source happening in the .NET space lately that may have been flying under your radar.

Related Links

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. I am a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web

NuGet Package of the Week: FluentAutomation for automated testing of Web Applications

March 4, '14 Comments [26] Posted in NuGet | NuGetPOW | Open Source
Sponsored By

FluentAutomation starting a testLast week I was exploring today's varied choices we have for Automated Browser Testing. There's headless WebKit "browsers" like PhantomJS and cloud powered multi-browser testing tools like BrowserStack and SauceLabs.

Selenium is kind of the gold standard and offers not only a lot of "drivers" but also a lot of language bindings with which drive a browser. Sometimes browsers update so fast there can be some version incompatibilities with Selenium, but for the most part it works great once you've settled in.

One option I've been looking at is FluentAutomation. It's a fluent automation API that supports Selenium as well as WatiN along with all their flavors and drivers. Since Fluient supports Selenium, that means you can use the Selenium ChromeDriver, IEDriver, Remote Web Driver or even the headless PhantomJS. FluentAutomation is on GitHub, of course, as well as on NuGet.

FluentAutomation has great (and growing) documentation and has adopted and interesting fluent style for it's API.

Now, not everyone likes a "fluent" API so it may take a while to get used to. Often you'll be doing things over many lines when it's really just one line, for example, this is one line:

I.Open("http://automation.apphb.com/forms")
.Select("Motorcycles").From(".liveExample tr select:eq(0)")
.Select(2).From(".liveExample tr select:eq(1)")
.Enter(6).In(".liveExample td.quantity input:eq(0)")
.Expect
.Text("$197.72").In(".liveExample tr span:eq(1)")
.Value(6).In(".liveExample td.quantity input:eq(0)");

Notice the method chaining as well as the use of CSS selectors.

FluentAutomation also has the cool concept of a PageObject to take your potentially brittle scripts and give them more structure. PageObjects group your actions, expectations, and assertions and let you reuse code when a page appears in multiple tests.

For example you could have a high level test (this is XUnit, but you can use whatever you want):

public class SampleTest : FluentTest {
public SampleTest() {
SeleniumWebDriver.Bootstrap(SeleniumWebDriver.Browser.Chrome);
}

[Fact]
public void SearchForFluentAutomation() {
new BingSearchPage(this)
.Go()
.Search("FluentAutomation")
.FindResultUrl("http://fluent.stirno.com/blog/FluentAutomation-scriptcs/");
}
}

Then you can have separate PageObjects that have your own public methods specific to that page, as well as assertions you can reuse.

public class BingSearchPage : PageObject<BingSearchPage> {
public BingSearchPage(FluentTest test) : base(test) {
Url = "http://bing.com/";
At = () => I.Expect.Exists(SearchInput);
}

public BingSearchResultsPage Search(string searchText) {
I.Enter(searchText).In(SearchInput);
I.Press("{ENTER}");
return this.Switch<BingSearchResultsPage>();
}

private const string SearchInput = "input[title='Enter your search term']";
}

public class BingSearchResultsPage : PageObject<BingSearchResultsPage> {
public BingSearchResultsPage(FluentTest test) : base(test) {
At = () => I.Expect.Exists(SearchResultsContainer);
}

public BingSearchResultsPage FindResultUrl(string url) {
I.Expect.Exists(string.Format(ResultUrlLink, url));
return this;
}

private const string SearchResultsContainer = "#b_results";
private const string ResultUrlLink = "a[href='{0}']";
}

You don't have to be all structure and OO if you don't want. You can just as easily write scripts with FluentAutomation and head in a different direction.

FluentAutomation along with ScriptCS = Automating your Browser with C# Script

I've usually used Python with my Selenium scripts. I like being able to just make a text file and start scripting, then run, debug, continue, all from the command line. It feels simple and lightweight. Creating a DLL and running Unit Tests in C# usually comes later, as I can move faster with a "scripting language."

You can do that with ScriptsCS as it gives you project-less C# that effectively is C# as scripting language. Combine this with FluentAutomation and you've potentially got the best of both worlds.

To install, first you need the Windows apt-get open source equivalent, the oddly-named and -spelled Chocolatey. Then you get ScriptCS and the packages for FluentAutomation.

  • Install Chocolatey - one line installation here
  • Run "cinst ScriptCS" from your command line to use Chocolatey to install ScriptCS
  • Now, get the ScriptCS script packages for FluentAutomation like this:
    • scriptcs -install FluentAutomation.SeleniumWebDriver
    • scriptcs -install ScriptCs.FluentAutomation

Now, as a quick test, create a folder and put a text file called start.csx in it with just these contents:

var Test = Require<F14N>()
.Init<FluentAutomation.SeleniumWebDriver>()
.Bootstrap("Chrome")
.Config(settings => {
// Easy access to FluentAutomation.Settings values
settings.DefaultWaitUntilTimeout = TimeSpan.FromSeconds(1);
});

Test.Run("Hello Google", I => {
I.Open(http://google.com);
});

Notice how there's no namespace, no classes, no main. It's just a script, except it's using C#. You can change the "Chrome" to "IE" or "Firefox" as well, to play around.

Random: I love this Selenium feature, exposed by FluentAutomation...take screenshot!

// Take Screenshot
I.TakeScreenshot("LoginScreen");

If you don't want ScriptCS, while it can act as a REPL itself, there is also the start of a dedicated FluentAutomation REPL (read–eval–print loop). This is basically a command prompt that lets you explore you app interactively and facilitates building your scripts. You can get the Repl as a Chocolatey package as well and just "cinst FluentAutomation.Repl"

You've got LOTS of choices in the world of automated testing. There's so many choices that there's just no good excuse. Pick a library, pick a language, and start automating your web app today.

Related Links


Sponsor: Big thanks to ComponentOne, a division of GrapeCity, for sponsoring the blog this week. Their widely popular .NET control suite, Studio Enterprise contains hundreds of data and UI controls such as grids, charts and reports that offer the functionality, features and support you need for current and future application development. Download your trial today!

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. I am a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web

Is the Windows user ready for apt-get?

May 28, '13 Comments [88] Posted in NuGet | Open Source | Tools
Sponsored By
Chocolatey installs Git

What it does

Chocolatey lets you install Windows applications quickly from the command line via a central catalog of installation scripts. You could install Git, 7Zip or even Microsoft Office (given a key.) The idea is seamless and quiet installations using a well-known key.

For example, once installed you can do this from and command line:

  • cinst git
  • cinst 7zip
  • cinst ruby
  • cinst vlc

That's basically it.

The catalog has grown so complete, in fact, that I recently wanted to install DosBox so I could play Zork. I took and chance and just "cinst dosbox" and it worked. THAT is a the promise that Chocolatey makes.

Getting Started with Chocolatey

You can get started by first installing the Chocolatey package manager. Copy paste this line to your command line and run it. (More on the fearfulness of this first step in a moment).

@powershell -NoProfile -ExecutionPolicy unrestricted -Command "iex ((new-object net.webclient).DownloadString('https://chocolatey.org/install.ps1'))" && SET PATH=%PATH%;%systemdrive%\chocolatey\bin

Presumably you like to know what command line stuff is going to do to your computer before you do it, so parse this line out. It's going to launch PowerShell to do the hard work. Nearly every Windows machine has PowerShell these days, and it's PowerShell that makes Chocolatey work.

Some folks have custom profiles so the -NoProfile switch suppresses custom profiles to prevent conflicts during installation. It launches a chunk of PowerShell script that it downloads from https://chocolatey.org/install.ps1/ then executes. Note that it's setting execution policy to unrestricted to do this. To be clear, it's executing code downloaded over the web, so there is a non-zero risk there. It then adds Chocolatey to your path (for this one prompt) so you can use it immediately. It'll be added to future instances of prompts automatically.

Look at https://chocolatey.org/install.ps1 now. It's a very clean and easy to read script. It downloads the Chocolatey installation zip file (which is actually a NuGet package), unzips it and continues the installation by running a scripts in the tools section of the package.

How it works

Chocolatey is a bootstrapper that uses PowerShell scripts and the NuGet packaging format to install apps for you. NuGet is the package management system that Windows Developers use to bring libraries down at the project level. Chocolatey (get it? Chocolatey Nu-Get?) extends that concept to bring applications down at the system level.

Today if you want to get 7Zip, you usually google for it, find the site, figure out the latest version or right version for your system, download it, run it, next next next finish and maybe add it to your path. Chocolatey does that for you.

Again, NuGet is libraries installed locally for projects, Chocolatey is applications installed globally for your whole system.

Chocolatey uses PowerShell scripts (that you never have to think about) that package developers use to chain installations and build dependency trees. Take the internals of a Git installation script for example:

try {
Install-ChocolateyPackage 'git.install' 'exe' '/VERYSILENT' 'http://msysgit.googlecode.com/files/Git-1.8.1.2-preview20130201.exe'

#------- ADDITIONAL SETUP -------#
$is64bit = (Get-WmiObject Win32_Processor).AddressWidth -eq 64
$programFiles = $env:programfiles
if ($is64bit) {$programFiles = ${env:ProgramFiles(x86)}}
$gitPath = Join-Path $programFiles 'Git\cmd'

Install-ChocolateyPath $gitPath 'user'

@"

Making GIT core.autocrlf false
"@ | Write-Host

#make GIT core.autocrlf false
& "$env:comspec" '/c git config --global core.autocrlf false'

Write-ChocolateySuccess 'git.install'
} catch {
Write-ChocolateyFailure 'git.install' $($_.Exception.Message)
throw
}

The most important part for you to take away here is the first line. Note that this Chocolatey script is downloading Git from the mSysGit Site. Chocolatey is not changing installers, making installers or hosting installers. It's automating the boring parts of getting software, but it's still getting that software from the same location as always.

Advanced Stuff

Once you learn the basics - and they're pretty basic - there's more depth to Chocolatey to explore. Beyond the cinst and cuninst there's other commands to make installing stuff on Windows easier. Remember, they're all in your PATH so you can call these commands anytime.

Each of these major sources can be called with cinst using the -source parameter like "cinst IISExpress - source WebPI" or using their own aliases for simplicity as shown below.

  • cwindowsfeatures - If you've ever opened Add/Remove programs then click Install Windows Features in order to setup IIS or Hyper-V then this command is for you. Some examples:
    • cwindowsfeatures  IIS-WebServerRole
    • cwindowsfeatures Microsoft-Hyper-V-All
    • cwindowsfeatures TelnetClient
      • Plus, you can always clist -source windowsfeatures for the complete list.
  • cwebpi - The Web Platform Installer is a great GUI for downloading any development tools you might need for Web Development on Windows. It's a catalog, an installer, and a chainer. There's also a command-line version of WebPI that Chocolatey integrates with so you can:
    • cwebpi IISExpress
    • cwebpi VWDOrVs11AzurePack_2_0
      • And again, clist -source webpi gets you a list of what you can do.

There's a more complete list at the Chocolatey Commands Reference including how it integrates with Cygwin, Gems and Python.

Security Issues

nugetlogoThis is a prickly one. How do you make a system that lets you install anything directly from the Internet quickly, cleanly, and easily without, well, installing something evil directly from the Internet? You'll want the communication with the server to be secure and the packages trusted, but you'll also want to make sure the packages haven't been tampered with since they were uploaded. There's the inevitable threat of a man-in-the-middle attack. You'll want to watch for malicious packages and enable quick takedowns if one sneaks by.

Security concerns aren't unique to Chocolatey, of course. They are a part of package repositories since their inception. The node npm repository had a security breach in March of 2012, and the folks at andyet explored the issues surrounding it, but also pointed out that personal responsibility has to have a role as well.

Linux's apt-get solves much of this with appropriate uses of cryptography and best practices that can (and should) be emulated. Packages in apt repos are signed with SecureApp, there are warnings if you're using a 3rd party repo or installing an unsigned package.

The Chocolatey team has been very quick to jump on security issues and they are very approachable. They've added SSL where appropriate and are aware of the work to come. If Chocolatey gets big (bandwidth and costs is a question in my mind) perhaps a non-profit organization would step in to help with not only costs, but also security audits and best practices.

Here's some points (edited for length by me) from a post from Chocolatey's lead, Rob in a post on their mailing list, also in march of 2012:

Security has a big future aspect of chocolatey. At the present I am the curator and I every day I get an email showing me all of the new packages that went in the day before. I look at all packages from new authors and I typically look at the first version of most new packages from authors I have good contacts with.

I've talked at length with others about having a moderated feed in the aspect of every package, every new version would be approved prior to showing up on the main feed. I am paying attention to how debian does things with multiple feeds and there are thoughts to move in that direction as well.

Security? In the future we are looking at a small group of folks be an approving body for nupkgs. We also talked about showing the hash for the nupkg, and possibly letting folks specify a hash for the installers so chocolatey can verify the things it downloads prior to execution.

Chocolatey's LIB folderCould I make a Chocolatey package called "FormatMyHardDrive?" Sure I could, just like I could ask you to open an admin prompt and format c: /q, but you won't, right? ;)

What's next?

Chocolatey is clearly not meant to be used by your "Gender Non-Specific Non-Technical Parent" and it does have some "competition" in the form of the Ninite GUI installation utility. While still not for the average Joe/Jane and having only a limited catalog, Ninite does fill a gap for the super-user to quickly get the common apps and utilities they want.

Additionally, is Chocolatey really apt-get? It's not installing libraries system-wide, although there's no reason it couldn't. Other open source projects like CoApp would like to be the Windows app-get although CoApp is more of a "system-wide libraries, C++ support, and Unix-like utilities" and Chocolatey is more of a "developer and poweruser utilities and their dependencies."

Chocolatey does install dependencies and you can see that happen yourself by trying out "cinst gitextensions" which itself has a dependency on git. Chocolatey will walk the graph and install what it needs before finally installing gitextensions.

Where Chocolatey, and ultimately Windows itself, falls down is with odd PATHing and install locations. Because Windows doesn't have formal install locations for things and because Chocolatey puts itself first in the PATH, it's possible to get one's self into odd situations where apps that were installed outside of Chocolatey don't line up with apps installed inside. For example, I installed Git with Chocolatey some months ago, then forgot about that version and installed a newer version of Git on my own. However, I kept hitting an old git bug because the Chocolatey version of Git was "first." I believe issues like this have changed with recent builds of Chocolatey, but the point remains: it's hard on Windows today to who installed what low-level utility, when, and where it ended up.

Branding

Now, by no means to I want to take away from the hard work done by Rob and the team, but (and I've said this to Rob before) I really have trouble getting past the name Chocolatey. Sure, there are two ways to spell "Chocolaty," which make it hard at least for me to type "Chocolatey" reliably. The -ey is a theoretically a valid variant spelling, but you can tell that that to the red squiggled underline in Word. But it's less the spelling and more the name itself. It lacks the nerdiness of an "npm," the gravitas of an "apt-get," or the poetic terseness of a "gem."  I realize that we are living in a world with companies called Hulu, Yahoo, Microsoft (seriously, MICRO SOFT, what is that?) and Google, but it's worth pointing out that a good name can really take a project to the next level. I'm not sure Chocolatey is the right name for this project, but that's 100% my opinion.

I encourage you, technical reader, to check out Chocolatey for yourself! It's a powerful tool, an engaged and growing community and an interesting piece of tech in its own right.

Is Chocolatey the apt-get Windows users have been waiting for? Sound off in the comments.


Sponsor: Big thanks to SoftFluent for sponsoring the feed this week! Check out their slick code generation tools: Less Plumbing, More Productivity! Generate rock-solid foundations for your .NET applications from Visual Studio and focus on what matters!

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. I am a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web

Project-less scripted C# with ScriptCS and Roslyn

April 24, '13 Comments [31] Posted in NuGet | NuGetPOW | Open Source | VS2012
Sponsored By
ScriptCS inside of SublimeText2 with the ScriptCS package giving SyntaxHighlighting

Glenn Block is working on something interesting that combines C#, NuGet, Roslyn (the new "compiler as a service") and his love of text editors and scripts. Now, with help from Justin Rusbatch (@jrusbatch) and Filip Wojcieszyn (@filip_woj) they are having all kinds of fun...using C# as a scripting language.

Every few years someone tries to turn C# into a competent scripting world, myself included. Often this has included batch files and MacGyver magic, file associations and hacks. Clearly the .NET community wants something like this, but we are collectively still trying to figure out what it should look like. PowerShell aficionados - and I count myself amongst them - might look at such efforts as a threat or a pale reinvention of PowerShell, but the fact remains that C# at the command line, be it as a script or a REPL, is an attractive concept.

Simply put by example, ScriptCS lets me do this:

C:\temp>copy con hello.csx
Console.WriteLine("Pants");
^Z
1 file(s) copied.

C:\temp>scriptcs hello.csx
Pants

That's Hello World. There's no namespace, no class, just some C# in a .csx file. Roslyn takes care of the compilation and the resulting code and .exe never hits the disk.

Self-hosting Web APIs

So that's interesting, but what about bootstrapping a web server using NancyFX to host a Web API?

Go and clone this repo:

git clone https://github.com/scriptcs/scriptcs-samples.git

Look in the Nancy folder. There's a packages.config. Just like a node.js application has a packages.json file with the dependencies in has, a .NET app usually has a packages.config with the name. In node, you type npm install to restore those packages from the main repository. Here I'll type scriptcs -install...

C:\temp\scriptcs-samples\nancy>scriptcs -install
Installing packages...
Installed: Nancy.Hosting.Self 0.16.1.0
Installed: Nancy.Bootstrappers.Autofac 0.16.1.0
Installed: Autofac 2.6.3.862
Installation successful.

Now, running start.csx fires up an instance of Nancy listening on localhost:1234. There's no IIS, no ASP.NET.

C:\temp\scriptcs-samples\nancy>scriptcs start.csx
Found assembly reference: Autofac.Configuration.dll
Found assembly reference: Autofac.dll
Found assembly reference: Nancy.Bootstrappers.Autofac.dll
Found assembly reference: Nancy.dll
Found assembly reference: Nancy.Hosting.Self.dll
Nancy is running at http://localhost:1234/
Press any key to end

There is also the notion of a "ScriptPack" such that you can Require<T> a library and hide a lot of the bootstrapping and complexity. For example, I could start up WebAPI after installing a Web API package that includes some starter code. Note this is all from the command line. I'm using "copy con file" to get started.

C:\temp\foo>scriptcs -install ScriptCs.WebApi
Installing packages...
Installed: ScriptCs.WebApi
Installation completed successfully.
...snip...
Added ScriptCs.WebApi, Version 0.1.0, .NET 4.5
Packages.config successfully created!

C:\temp\foo>copy con start.csx
public class TestController : ApiController {
public string Get() {
return "Hello world!";
}
}

var webApi = Require<WebApi>();
var server = webApi.CreateServer("http://localhost:8080");
server.OpenAsync().Wait();

Console.WriteLine("Listening...");
Console.ReadKey();
server.CloseAsync().Wait();
^Z
1 file(s) copied.

C:\temp\foo>scriptcs start.csx
Found assembly reference: Newtonsoft.Json.dll
...snip...
Listening...

Pretty slick. Add in a little Live Reload-style action and we could have a very node-ish experience, all from the command line and from within your text editor of choice, except using C#.

Note that this is all using the same CLR and .NET that you've already got, running at full speed. Only the compilation is handled differently to give this script-like feel.

Installing ScriptCS

The easiest way to install and use ScriptCS is to use Chocolatey (a system-wide NuGet-based application/component installer. "Chocolatey NuGet," get it?) And yes, it's Chocolatey spelled incorrectly with an "-ey."

You can use Chocolatey to do things like "cinst 7zip" or "cinst git" but we'll be using it just to get ScriptCS set up. It's also easily removed if it freaks you out and it installs no services and won't change anything major up save your PATH.

First paste this into a cmd.exe prompt:

@powershell -NoProfile -ExecutionPolicy unrestricted -Command "iex ((new-object net.webclient).DownloadString('https://chocolatey.org/install.ps1'))" && SET PATH=%PATH%;%systemdrive%\chocolatey\bin

This will PowerShell, run https://chocolatey.org/install.ps1 and add Chocolatey to your path.

Then, run

cinst ScriptCS

Which will put ScriptCS in a path like C:\Chocolatey\lib\ScriptCs.0.0.0 while Chocolatey makes it available in your PATH.

Sublime Text or Visual Studio

You can get syntax highlighting for your CSX files inside of Sublime Text 2 with the "ScriptCS" package you can install from package control. If you're using Visual Studio you can get the Roslyn CTP to turn on CSX syntax highlighting.

You can use PackageControl in SublimeText2 and install the ScriptCS package

You can even debug your running ScriptCS projects by opening the ScriptCS.exe as a project. (Did you know you can open an EXE as a project?) Add the .csx script to the command line via Project Properties, drag in the scripts you're working on and debug away.

Debugging requires the Roslyn SDK, although personally, I've been doing just fine with scripts at the command line which requires nothing more than the basic install and a text editor.

It's not clear where ScriptCS is going, but it'll be interesting to see! Go get involved at scriptcs.net. This kind of stuff gets me excited about the prospect of a compiler as a service, and also cements my appreciation of C# as my enabling language of choice. Between C# and JavaScript, you can really get a lot done, pretty much anywhere.

I'll have a video walkthrough on how this works as I explain it to Rob Conery up on TekPub soon! (Here's a referral coupon for 20% off of Tekpub!)

What do you think?

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. I am a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web

NuGet Package of the Week #13 - Portable HttpClient makes portable libraries more useful

March 21, '13 Comments [17] Posted in NuGet | NuGetPOW | Win8 | Windows Client | WinPhone
Sponsored By

Reference Assemblies include .NET Portable AssembliesWhen you've got an idea for an app, it's likely that you've got the idea for that app in more than one place. By this I mean, you'll start with a phone app, then make a desktop app, then a web app. Or you'll make a game on one platform and then want it to work anywhere. In fact, with the rise of Xamarin, C# lets you put an app in every AppStore in the world with one language.

You likely already knew that you can target different versions of the .NET framework. You likely also know that there are small .NET Frameworks like Silverlight and the tiny .NET Micro Framework.

You can also target XBox and Windows Phone, OR better yet, target a profile called Portable Libraries that I've briefly mentioned before. Portable Libraries are a great idea that have some issues when you try to really use them. There's actually a great (if a little older) video with the inventors over at Channel 9. Note that Portable Libraries ship with Visual Studio 2012 and are a supported thing.

The idea is that you write a library that contains as much shared functionality as possible and then every application uses your now "portable" library. However, the subset of classes that are available are a subset. That means you can only use things that are available in the intersection of the targets you choose. Check this dialog:

Choose your target framework

And check out the Supported Features table at this MSDN article on Portable Libraries to find out what you can use where. Here's a graphical table I stole from Daniel.

PortableLIbraryChart

However, most folks that use Portable Libraries have ended up using them mostly for ViewModels - just simple classes without any real functionality. Almost as if we had a DLL full of structs. There are some great posts on how to make Portable Class Libraries work for you using architectural techniques like indirection and appropriate interfaces.

The number one complaint around code resuse and the number one voted item over at the Visual Studio UserVoice was "Add HttpClient support in Portable Class Libraries (including Windows Phone 8)." Why? Because the GETting of a remote resource via HTTP is such a fundamental thing that it'd be awesome to be able to bake that data access into a portable library and use it everywhere.

Now there is a Portable Http Client and you can get it via NuGet!

install-package Microsoft.net.http -pre

Here's an example of what the code looks like for a GET. Note that I'm using async and await also.

public static async Task<HttpResponseMessage> GetTheGoodStuff() 
{
var httpClient = new HttpClient();
HttpRequestMessage request = new HttpRequestMessage(HttpMethod.Get, "http://hanselman.com/blog/");
var response = await httpClient.SendAsync(request);
return response;
}

.NET Portable Subset

If you were going to make a Twitter client (lame example, but bear with me) you could now put the JSON HTTP data access code in one library and share it across Windows Phone, Windows Store, WinForms, Console, whatever.

I'm hoping that the folks at MS and the folks at Mono will continue to work to make Portable Libraries a good option for Mono as well. I've been advocating (and pushing) to make something happen as well, as have the Portable Libraries folks. You'll find lots of working in the space around the web, so fear not, code reuse, either through Portable Libraries or via linked code files at compilation time is deeply possible. The game "Draw A Stickman Epic" achieved 95% code reuse by writing the game in C# with MonoGame!

.NET 4 or Windows Phone 7.5

If you want to use this HttpClient on .NET 4 or Windows Phone 7.5, note you might get a compile error if you use async and await.

Cannot await System.Threading.Task<HttpRequestMessage>

This is because .Net 4.0 and Windows Phone 7.5 did not support the async/await keywords. In order to fix this add a reference to the Microsoft.Bcl.Async nuget package, which adds the support for async and await in .NET 4 and WP7.5. Here's a post with more details on how this backport works.

Related Links

About Scott

Scott Hanselman is a former professor, former Chief Architect in finance, now speaker, consultant, father, diabetic, and Microsoft employee. I am a failed stand-up comic, a cornrower, and a book author.

facebook twitter subscribe
About   Newsletter
Sponsored By
Hosting By
Dedicated Windows Server Hosting by ORCS Web
Page 1 of 6 in the NuGet category Next Page

Disclaimer: The opinions expressed herein are my own personal opinions and do not represent my employer's view in any way.