Platform Considerations
The following information encapsulates much of my thinking regarding how to power this site. It is based on experimentation and observations as of 2015-05-02.
Architectural/Philosophical Considerations
Static Site Generation - I didn’t want a full CMS like Drupal, a hybrid CMS like WordPress or even a static file CMS like Kirby or Pico. I wanted something with a minimum amount of overhead that would allow me to host static files in places like GitHub Pages. Ultimately, StaticGen maintains one of the more interesting lists of static generators, so if you want a comprehensive view of the space, look there, but listed below are thoughts and observations from my own foray into this space.
Language - While there was some temptation to scratch an itch with regard to learning new languages, I wasn’t intrigued with the prospects of going overboard there. I like the idea of using node.js – a lot. I used to be intrigued with Python, but decided against it for this project. Despite being very popular in this space, I decided to avoid Ruby. I wanted to avoid anything based on .Net too. Oddly, I couldn’t find much intriguing in the PHP space, but would have found that workable.
Data - Ultimately, I want a tool that will allow me to augment simple site markdown files with data from an array of sources. I could just render everything down to markdown or html files, but I wanted to use data sources that would allow me to more flexibly manipulate what gets rendered. JSON was a requirement for me, but I’m becoming a fan of YAML and TOML. They are a bit more conveninent to work with in some situations and I may consider switching to one of them for some data sources I create.
Features - Basic support for themeing, templating, and partials are requirements for me. Live reload is a “nice to have” feature, but not that important in this case. It could help for some iteration, but given the scale I hope to operate at, I’m resigned to the belief that it isn’t going to be practical for use on the site I’m developing. The prospects for leveraging SASS/LESS are very attractive, but aren’t a defining characteristic for me. Pagination and classification (through tagging and/or taxonomy) were some of the key features I hoped to leverage.
Performance - I need something sufficently performant. Knowing the end-product (the rendered site) would be fast, throughput for site generation would be a luxury, an added convenience, but not a defining characteristic.
Open Source - I didn’t want to use any site generator that wasn’t open source even if it was free/inexpensive. I hope to find ways to contribute back to the systems I leverage.
Platform Summary
My search primarily focused on the following platforms. I’ve highlighted a few of the features that were key to my own consideration. I ruled out Jekyll due to its dependency on Ruby, but included it here due it is prominence in the marketplace.
| Tech | Language | Status | Tagging | Pagination | Data |
|---|---|---|---|---|---|
| Harp | Node.js | Stalled | Unsupported | Unsupported | Potential |
| Hexo | Node.js | Active | Supported | Supported | Potential |
| Hugo | Go | Active | Supported | Supported | Potential |
| Jekyll | Ruby | Active | Supported | Supported | Unknown |
Harp
I really like Harp, but consider it a last resort option at this stage. More than anything else, the need to host a metadata file in json to generate the site isn’t attractive for the scale I want to use for this project. For smaller sites that don’t need pagination and classification, it would be great.
| Feature | Status | Notes |
|---|---|---|
| Status | Stalled | Activity has waned as of 2015/05/02 |
| Pagination | Unsupported | It doesn't natively support pagination. I could find alternate routes by using data files, but the level of effort would be signficant. |
| Tagging | Unsupported | A reasonable path forward could be found in the use of data files |
| Data | Possible? | My needs could be supoprted w/ data files per "node", but haven't investigated it at this stage. |
Hexo
I really wanted to like Hexo, but abandoned it very quickly. I couldn’t handle the data files that I’d already tried with Hugo, so I decided to skip investigating it more deeply.
| Feature | Status | Notes |
|---|---|---|
| Status | Active | It seemed to be reasonably active |
| Pagination | Supported | It natively supports pagination, but didn't explore it deeply. |
| Tagging | Supported | It natively supports tagging, but didn't explore it deeply. |
| Data | Possible? | It natively supports use of secondary data files, but after attempting to use it with a large collection of data, Hexo proved unresponsive, so I abandoned. |
Hugo
Hugo is very intriguing. Before hitting some boundaries with data, it was extremely fast. The only thing missing from my wish list right now is SASS/LESS support and I think that’s on the backlog. The Go Templating system is a little foreign to me, but I’ve had no problems navigating it so far. EJS was far more familiar to me, but given everything else it does well, this is a reasonable trade-off. With node.js based systems, my learning curve for contributing back would be much smaller. If I continue progressing forward with Hugo, I’ll have to consider other ways to contribute, but I may break down and immerse myself more deeply in Go if Hugo proves a durable solution for me.
| Feature | Status | Notes |
|---|---|---|
| Status | Active | It seemed to be reasonably active and the support channel was quite responsive for the one issue I'd submitted. |
| Pagination | Supported | It natively supports pagination. |
| Tagging | Supported | It natively supports tagging. I'm still working through it. Not sure if pagination is supported at the tag level yet. |
| Data | Possible? | It natively supports use of secondary data files. It eventually collapses under the weight of large amounts of data. See my bug report at https://github.com/spf13/hugo/issues/1065 for more information. I may try switching to a Linux VM to see if results vary there, but it didn't sound promising. |
Jekyll
Jekyll is one of the most popular static site generators available. Ultimately, I didn’t want to use a Ruby based system, so I didn’t pursue it.
Conclusion
At this stage, I’m pressing forward with Hugo, but I’ll continue to keep an eye out on alternative platforms. Until something else captures my attention, I think it is well suited to drive the site if I make some sacrifices on the data to work around the bug I highlighted above.