BoxLang recently released it’s 1.4 version, and one of the cooler parts of that update was many improvements to MiniServer. MiniServer is a lightweight web server that makes it easy to spin up and test BoxLang web applications.

Updates in the last version included automatic .env loading (which is coming soon to the boxlang CLI as well), websocket support, health checks, and more, but the one I care the most about is URL Rewriting support.

Rewrite support is fairly simple. To turn it on, pass --rewrites when running boxlang-miniserver. This will use the file index.bxm for any request that doesn’t match a file (no matter what the extension). You can also specify a particular file as well: boxlang-miniserver --rewrites router.bxs.

When you running MiniServer this way and make a request for something that can’t be found, your file will run and you can inspect the request (typically via cgi.path_info) to decide what to do. Here’s an example from the docs:

switch( cgi.path_info ) {     case "/":         // Home page         include "views/home.bxm";         break;      case "/products":         // Products listing         include "views/products.bxm";         break;      case "/products/":         // Individual product (extract ID from URL)         productId = listLast( cgi.path_info, "/" );         request.productId = productId;         include "views/product-detail.bxm";         break;      default:         // 404 page 		bx:header statusCode=404;         include "views/404.bxm"; } 

I think the most interesting part of that example is the third one where you can enable clean URLs in the form of, /product/500 or /product/999. For the most part, this all "just" works, but I thought I’d kick the tires a bit and build out two demos.

Building a Mini Blog Server with MiniServer

I’ve already built a simple blog in BoxLang earlier this year when I was first learning the platform. I thought it would be interesting to see if I could get BoxLang to grok my ‘real’ blog source files and URL structure. My blog uses 11ty which is a powerful state site generator built in Node. I’ve configured my blog to look for blog posts under posts folder. As I’ve got way too many blog posts, I’ve organized them by year, month, and day.

Each blog consists of front-matter on top, and excellent enterprise-grade content in Markdown. Here’s an example from the post I wrote this weekend:

Looking at the front matter, you can see the permalink option is how I define my URLs here. As a test, I took a year’s worth of posts, copied it under a new folder in a directory called posts, and whipped up a simple BoxLang class to handle:

• Recursively reading in all the Markdown files under it.
• Find the front matter and parse it with the BoxLang Yaml module.
• Render the Markdown using the BoxLang Markdown module.
• Support returning all posts as well as being able to look up a post via it’s permalink.

Here’s the class in question, and keep in mind I wrote this incredibly quickly and just to server the demo:

class {  	property name="postDirectory" type="string";  	private function fmParse(s) { 		data = {}; 		fm = s.reFind('---(.*?)---(.*)', 1, true); 		/* 		I'm a bit rusty on reFind, but fm.match will be an array where [2] is the str I want. 		I'm not 100% confident of this 		*/ 		//writedump(fm); 		if(fm.match.len() != 3) { 			return { data:{}, contents:markdown(s) }; 		} 		 		data.append(yamlDeserialize(fm.match[2]));  		content = fm.match[3];  		return { data:data, contents:markdown(content) } 	}  	public function getPosts() {  		blogCache = cache(); 		return blogCache.getOrSet('postCache', () => { 			println('not in cache'); 			posts = []; 			postFiles = directoryList(path=variables.postDirectory, recurse=true, filter="*.md");  			postFiles.each(p => { 				contents = fileRead(p); 				post = fmParse(contents); 				// parse .date to a full date 				post.data.date = parseDateTime(post.data.date); 				posts.append(post); 			});  			// do a quick date sort 			posts.sort(function(a,b) { 				return dateCompare(b.data.date, a.data.date); 			});  			return posts; 		}, 60 * 60);  	}  	public function findPostByPermalink(permalink) { 		posts = getPosts(); 		match = posts.find(p => { 			return p.data.permalink === permalink; 		}); 		if(match === 0) return; 		else return posts[match];  	} } 

While BoxLang had no trouble finding and parsing my files really quickly, I did go ahead and add an hour cache to the getPosts method to make it instantaneous.

Cool, now let’s use it. First, I set up a quick Application.bx file:

class { 	this.name = "rewrite_blog_demo";  	public function onApplicationStart() { 		application.blogService = new blogService(postDirectory=expandPath('./posts')); 	}  	public function onRequestStart() { 		// remove me 		application.blogService = new blogService(postDirectory=expandPath('./posts')); 	} } 

Yes, I left the "remove me" block in there. I was going to remove it before posting, but figured I’d share the hack I did to make development a bit easier. BoxLang web applications support an applicationStop method you can use for a ‘more proper’ way to restart an application, but for this simple test, this particular hack was fine. Ok, now to the home page:

<bx:script> posts = application.blogService.getPosts().slice(1,10);  </bx:script>  <h2>Blog Posts</h2>  <bx:loop item="post" array="#posts#"> 	<bx:output> 	<p> 	<a href="#post.data.permalink#">#post.data.title#</a> (#post.data.date.dateFormat('short')#) 	</p> 	</bx:output> </bx:loop> 

I begin by getting my posts and slicing it to the top 10. I then loop over each, print out the title and date, and link to the permalink. Here’s a subset of the output HTML:

<h2>Blog Posts</h2> <p> <a href="/2025/08/07/integrating-location-data-with-built-in-chrome-ai-for-better-image-insights">Integrating Location Data with Built-in Chrome AI for Better Image Insights</a> (8/7/25) </p> <p> <a href="/2025/08/05/building-a-comic-book-reader-in-boxlang">Building a Comic Book Reader in BoxLang</a> (8/5/25) </p> 

Alright, let’s look at the rewriter template, rewriter.bxs:

/* Ok, so technically, we should validate the path, ensure it matches /YYYY/DD/MM/slug,  but for now I'm just going to assume it does */ post = application.blogService.findPostByPermalink(cgi.path_info); if(post) { 	request.post = post; 	bx:include template="post.bxm"; } else bx:location url="/"; 

As the comment says, I’m kinda assuming I’m only running for blog post requests. The cgi.path_info value will match exactly the permalink and so the findPostByPermalink should return the right data, and if not, I simply redirect back.

A better approach would be to sniff for a permalink type structure and redirect on a bad one, and then use a 404 header result for other responses.

The final bit is just my basic post template:

<bx:output> <p><a href="/">Home</a></p> <h2>#request.post.data.title#</h2>  #request.post.contents# </bx:output> 

As I said, basic, but you get the idea. Check out the full demo on the BoxLang demos repo. Now let’s kick it up a notch!

Cloudinary in BoxLang

I’ve been a huge fan of Cloudinary for years now and use it in multiple places on this blog. They provide an incredibly powerful, URL-based API for image and video transformations. By that I mean you can take your image tag, redirect it to Cloudinary and by adding stuff to the end of the URL, perform a near infinite number of transformations on images.

So naturally I thought – lets rebuild that in BoxLang!

Now to be clear, what I ended up building isn’t even 1% of what Cloudinary supports. It was really just a test to see what I could do with rewrites and the BoxLang image module.

I began with a simple idea. You would request an image in the web app, and if you added X/Y to the end of the path, the image would be resized to fit that bounding box. (Initially I was going to support passing a width only, but this exposed a bug in the image module you’ll see mentioned in my code.) I then went a step further and added support for 3 effects: grayscale, blur, and negative.

Here’s the rewriter script I built:

 if(cgi.path_info.findNoCase('/img/') === 1) { 	imgPath = cgi.path_info.replaceNoCase('/img/',''); 	/* 	formats are: 	file  	file/w (change to width of w - not working - https://ortussolutions.atlassian.net/browse/BLMODULES-81 ) 	file/w/h (change to width of w and height of h) 	file/w/h/effect where effect is one of: grayscale/blur/negative 	*/ 	parts = imgPath.split('/'); 	img = expandPath('./source/') & parts[1];  	if(!fileExists(img)) { 		println('file didnt exist'); 		bx:header statusCode=404;  		return; 	}  	img = imageRead(img);  	if(parts.length >= 4) { 		effect = parts[4]; 		if(effect === "grayscale") img.grayscale(); 		if(effect === "blur") img.blur(20); 		if(effect === "negative") img.negative(); 	}  	if(parts.length >= 3) { 		img.scaleToFit(parts[2], parts[3]); 	} else if(parts.length == 2) { 		// unfortunately, buggy right now 		img.scaleToFit(parts[2]); 	} else if(parts.length == 1) { 		// another small bug, if you don't do _something_ with the image, getBlob fails. Bug is filed. 		img.info(); 	}  	bx:header name="Content-Type" value="image/jpeg";  	bx:content variable="#img.getBlob()#";  } 

As you can see, I basically load up the image in a source directory, perform the relevant transformations, and return the binary data. While working on the blog post I uncovered another issue with the image module which is corrected with the img.info() throwaway line. This issue basically relates to the image now returning the right data unless I did something to it, hence the info call that does nothing.

Ok, so how is it used? Here’s two examples:

<p> <img src="/img/20250106_073555.jpg/500/500"> </p> <p> <img src="/img/20250106_073555.jpg/500/500/grayscale"> </p> 

And here’s the result:

Pretty snazzy, right? I could also do some caching here to improve performance, but as I said, this was just a proof of concept. You can find the complete demo here: https://github.com/ortus-boxlang/bx-demos/tree/master/webapps/rewritedemo/imgdemo

As always, let me know what you think and leave me a comment below. When I don’t get comments I believe I may have died without knowing it and I’m just a ghost. I get the same feeling when the faucets in the bathroom don’t turn on. Freaky, I know.

Photo by Maria Lupan on Unsplash

Raymond Camden