Yep, you read that right. I’m building tools in Go.
I can hear you now…. But Wes you are a SQL Server guy, a WINDOWS guy! What about C#? What about Powershell? What about ANY THING ELSE???
Sometimes you make choices, not based on what you want to use but what you must use when building your tool set.
Here was my decision making path I hope it isn’t too complex.
It must run on everything from Windows Server 2003 through Windows 2012 R2 where I don’t have total control over the OS or what is installed on it.
That was it. I had no clue it would be so difficult to accomplish. Let me run down several candidates.
This is my default go to stack. I have been writing in C# for quite a while and I’m fairly comfortable with the language and love the amount of third party libraries and tooling around it. Oh and ReSharper. Microsoft is also pushing hard to open up C# and all aspects of .Net going so far as buying Xamarin, the developers behind mono. All of that is great! There is still one HUGE downside with the .Net framework. You have to have the .Net framework.
But Wes, the framework has been shipping for YEARS and is already installed on probably 90% of all the servers you are going to be using your tools on! You are absolutely correct, which version though? That is a horrible problem to solve. Pretty much the only way I found to solve it was to only use the 2.0 framework. That meant a HUGE part of the improvements to the .Net framework was now unavailable. Oh yeah, ever try to file a bug report to a third party library developer about a bug that only effects the 2.0 framework? You end up building “new” code and now maintaining legacy code from third parties.
Not all is lost! You can compile a “native app” but that can only target Windows 10, unless you use mono. With mono It is possible to build a stand alone self contained executable for Windows.
Well…. The up side is I get the new shiny bits but mono doesn’t implement the full .Net framework. Also it is hit or miss on third party libraries and if they will also work with mono. Oh and the build process is right out of the 90’s. If all of that wasn’t enough there are possible licensing issues with embedding the mono run time in your application vs. linking to it as a dynamic library.
No seriously, it suffers from the same kind of problems. I am back to building scripts that target Powershell V1 and hoping security issues don’t blow up in my face.
I don’t want the Ruby or Python folks to flatten my tires, I’m not saying they are the same weakly typed, single-threaded interpreted language. Python doesn’t have Rails. ZING!
All kidding aside, I personally like Python over Ruby but that again is a personal preference. Honestly though I am more old school and really love strictly typed languages.
Both have TONS of third party libraries, that run on Linux or OSX. Windows is a bit thinner though. They both have ways of rapping up their respective run time into an object that looks like a standalone executable. There are several companies running both Ruby and Python on Windows and ship software installers to end users. I know it can be done. I also know it is kinda painful too.
I didn’t stop there!
Lua, another favorite language of mine mostly due to is simplicity and ability to write and then read that same code again later.
Open Pascal/Delphi was given a cursory look but it just doesn’t have the critical mass it once had.
C++ is awesome if you want to crash both your program and the computer. I dare you to figure out what part of the 1000 page specification you aren’t going to use to maintain some level of productivity.
C has a huge draw for me but again it is a double-barrel shotgun but one barrel points backwards.
I eventually found my way to Googles’ Go language. I had poked around with it several years ago but it was a pain to compile for Windows. Fast forward to 2016 and now Go compiles easily on Windows. Most third party packages are written in Go which means it generally compiles on Windows with zero changes. These executables are statically linked that means executables that will probably be larger than some but come with the bonus of not having to carry around all the run time and libraries or hope they are installed on the machine.
Go, as a language, is simple. It isn’t as terse as say K but it is much closer to C and Python than C# or C++ is.
Go is absolutely opinionated. There is the way I like to write code then there is the Go way. They don’t always agree. You want generics? NOT IN MY LANGUAGE YOU DON’T! Declare a variable and then don’t use it? No compile for you! Want to pick your own way of structuring your source code files? That ain’t going to fly.
These are all choices specifically made to satisfy writing LOTS of code worked on by LOTS of people, a Google’s worth so to speak.
So, why does this matter to me and the hand full of people that may actually have to work on my code?
More than you think. There are suggestions, fences, guide rails and walls in the Go ecosystem. They are there to help prevent bugs and improve productivity at scale. I have worked with a lot of programming languages over the years. My favorite ones allowed me to build stuff quickly and hand it to someone else knowing they could also get up to speed quickly. To that end I am finding that I agree with some of the guiding principles of Go.
Simplicity, safety and readability are paramount.
Minimal design: There is one way to write a piece of code.
It’s about expressing algorithms, not the type system.
Tooling is as important as the language.
Build times shouldn’t take over night.
Compiles naively across all platforms.
Things of interest should be easy; even if that means not everything is possible.
That means I have to give up some control. It also means Go may not be the best choice in all situations. To me Go is starting to feel like T-SQL, it isn’t horribly complicated yet tremendously powerful. I’ll be writing about the good and the bad around Go, or golang if you like, soon. I’m only in my first two weeks with it now.