Friday, October 31, 2008

LINQ to SQL death rumors


I think it all started with this post from Tim Mallalieu reporting some directions for LINQ to SQL and the ADO.NET Entity Framework. Ayende and others have had some strong reactions (LINQ to SQL is dead) and there's now a StackOverflow thread you can use to follow the discussion.

It will be interesting to track over the next few days what other responses pop up. A few things that come to my mind:

1. This is an opening for people using Linq to SQL to migrate to SubSonic or NHibernate. I've been using SubSonic lately and enjoying myself quite a bit.

2. A big disappointment for people like Benjamin Day who I saw speak at VSLive and who have incoprorated LINQ to SQL into their development projects. Next time wait for Microsoft to 2.0 something before spending too much time working it into your dev cycle.

3. I wonder if Microsoft hasn't thought of spinning off some of the tool makers as different companies. Perhaps there is an MBA out there who could draw some charts and convince management to have ADO.NET Entity Framework and the LINQ to SQL folks compete/copy in the same space. If the team isn't fully formed, why not make one? And then how cool would that be: profit centers that encourage developers to use Microsoft tools and competition to get the best ideas out the door. The trade off of Balkanization versus One True Product is that the former will mean that cool new ideas belong to you - the company won't have to get the ideas from outside sources. 

4. I wonder if we aren't moving towards a tipping point of Microsoft developers "getting" open source and shifting their tool/framework usage outside of Microsoft. I know lots of people who won't use __anything__ unless it has a Microsoft logo because it feels safer and more commoditized. Moves like this seem to go against that since open source projects tend to live a longer time and technologies / frameworks from companies become "obsolete" (how many versions of ADO can you count?) because of a product cycle that requires new purchases every 2 years or so. I would argue with my boss on a new project that it's safer to use NHibernate than ADO.NET Entity Framework if this indicates a chance that the Entity Framework could be the next LINQ to SQL project.


Thursday, October 23, 2008

.NET Survey


Scott Hanselman did a survey of .NET usage and posted results on his blog. A few interesting notes:

1. Heavy use of older stuff: ASMX, DataSet, WebForms, Windows Forms
You'd be hard pressed to find new blog entries on it but it shows that older (tried) technologies don't sync up with what's popular. The last few projects I've worked on have had development cycles that far exceed the excitement of their underpinning technology. One needs to remember that excitement and hype aren't what make something useful, it's use. So the people who talk down DataSets do so in the face of what obviously works for a silent majority who are busy getting stuff done.

2. Some new stuff won't take, some new stuff takes off
The respondents using CardSpace and WorkFlow are dismal. I can match the "unscientific" survey with a lot of my own experience in the last few years and conclude it's within a broader trend. On the other hand, LINQ To SQL has very heavy use for something relatively new, almost on par with ADO DataSets which one could argue are in the same competitive space. That says to me that in some cases adoption is just slow when there's a "legacy" counterpart for a technology but in others, when the new technology is compelling, people take to it very quickly.

3. WinForms > WPF
More than twice the respondents but here's my conclusion: it's all about the toolset, not the technology. WPFs tools are not on par with their WinForms equivalents. The dual Blend + VS strategy doesn't make sense for a large majority of us who don't have full time designers we work with and the lack of built in controls (such as a DataGridView) mean as cool as rounded edges are, function trumps form when programming.

4. MVC?
Last I checked this was *just* released as a beta. More people claim to use that than WPF which seems to indicate a Hanselman type majority in respondents. Even if I wanted to jump into MVC, until it's at a "1.0" release I couldn't (and I'm wondering how others seem to be doing so) justify starting or migrating a major product to a product/framework that isn't yet complete.


Wednesday, October 22, 2008

Recursive Copy ala LINQ


Tonight I needed to copy a bunch of images, no matter their level in the hierarchy to a single folder. Cake walk with C# and LINQ, I'll post the Python equivalent tomorrow.

string sourceDir = @"C:\thesource";
string targDir = @"c:\thetarget";
var files = new DirectoryInfo(sourceDir).GetFiles("*", SearchOption.AllDirectories)
.Where(p=> !p.Extension.Contains("db")) // exclude the pesky windows db file
.Select(p => p.FullName);
foreach (string path in files) {
File.Copy(path, Path.Combine(targDir, Path.GetFileName(path)));



Monday, October 20, 2008

Better Know a Framework: Zip / Package Multiple Files in C#


Making a single archive out of multiple files in .NET is easy once you know where to look but I spent far too long realizing there is a System.IO.Packaging namespace (not in the mscorlib, but you need to reference WindowsBase assembly).

// target file
string fileName = @"C:\temp\";

// point to some files, say JPEG - did I say I love LINQ?
var filesToZip = new DirectoryInfo(@"c:\temp\").GetFiles("*.jpg").Select(p => p.FullName);

using (Package exportPackage = Package.Open(fileName))
foreach (var file in filesToZip)
using (FileStream fileStream = File.OpenRead(file))
Uri partUriDocument = PackUriHelper.CreatePartUri(new Uri(Path.GetFileName(file), UriKind.Relative));
PackagePart packagePart = exportPackage.CreatePart(partUriDocument, System.Net.Mime.MediaTypeNames.Image.Jpeg);
CopyStream(fileStream, packagePart.GetStream());

// somewhere else, the CopyStream
private static void CopyStream(Stream source, Stream target)
const int bufSize = 0x1000;
byte[] buf = new byte[bufSize];
int bytesRead = 0;
while ((bytesRead = source.Read(buf, 0, bufSize)) > 0)
target.Write(buf, 0, bytesRead);

The trap in looking for this was that searches for this all point towards the GZipStream which is nice for a single file but not applicable for packaging multiple files. In my case I wasn't even interested in compression, just a single archive. You'll find additional documentation for the library here.

I'm half tempted to write a nifty tool in IronPython for command line packaging but in a world with 7-zip, that's simply foolish.


Saturday, October 18, 2008

Mashing Up: JQuery and Windows Forms, a first WIB


A lot of people are down on Windows Forms as something passe. I have a feeling there are a lot of people like me who not only must use it for practical reasons but also think there are advantages to fat clients that sit on the Windows platform. I'm getting to a point where I will start doing more than poking around with WPF but at present two factors hold me at bay: the control library is not robust enough (another tutorial of how to make a custom button in blend isn't incentive enough) and XAML is still a bit daunting for what usually end up being rather involved UIs. Here, for example, is a UI from an application I've been working on for a while (small intentionally to impart an idea of complexity without revealing any details):

I think a lot of "do it in XAML" problems can be solved with a more web development style paradigm in Windows. A simple example is template driven databinding with styles. That's easy to do in web applications but a little less flexible in a control based world (think a GridView with CSS versus the Windows Forms DataGridView).

As a proof of concept I decided to use the System.Windows.Forms.WebBrowser in a Windows Forms application to get cool features from JQuery while maintaining the strengths of a windows app. Here's the scenario: a photo manager application. It builds and manages a database of photos from a digital camera. Rather than uploading the 300+ pictures taken in a single session to Flickr (or in conjunction with that act) one can open it up, start annotating their files and saving the comments to a portable "database" that they can back up along with the files on their home machine.  On any machine with the application they point to the directory and can view their original comments along with the pictures. Here is a screen shot:

The strength in this scenario of Windows Forms is the ease with which one can build the editing interface: textboxes, buttons and other "drag / drop" pieces of the control library. The strength of JQuery is in the display - I've chosen to use an animation and the accordion to spruce up the interface where the user looks at their photos.

You can download the whole thing here. (Very 0.1 but posted as an example/idea - just use the next / previous buttons to view some photos, get a feel of things).

I wonder if this paradigm can be a kind of Ajax "aha" moment; Ajax was centered on the use of a single, relatively simple MSXML component. Though System.Web.Forms.WebBrowser is not simple, it's a single component that has a lot of room for leverage.  Ajax set up some pretty standard protocols / approaches to using asynchronous HTTP requests from a client - I think an interesting parallel could exist in wrapping jQuery and Html controls into a kind of framework that is easy for any developer to use with the WebBrowser control. Imagine a grid you drop with some properties and a CSS.  The last thing that made Ajax pop was the catchy name. I love naming things myself so I've been calling this kind of application a WIB (Windows + Web) but perhaps there's something more catchy one could use.

Just some ideas, I'd be curious in the following:

1. Any existing applications written with this approach.
2. Any potential applications that could benefit with the pros of both Windows and Web
3. Any glaring shortfalls



ps. Thanks to Corey for encouragement on the blogging tip.

Wednesday, October 01, 2008

YUI File Disjoin Complexity HELP


I got a little help yesterday from Dav Glass and Eric Miraglia for my issue on selection of YUI javascript files. What I'd had distributed over a bunch of files is now:

<script type="text/javascript" src=""></script> 

Hm... a bit better. Both Dav and Eric confirm that the canonical solution to this problem is to spend some quality time at the YUI Configuration Site which will allow you to pick the widgets and utilities you've leveraged in order to combine it into a more simplified call.

A someday project: the following style of implementation to make the reference a bit more crisp (I didn't add everything from above, but the pattern should be easy to observe):

<script type="text/javascript" src=""></script> 


Anyhow, thanks for the help.