Any tips for large, resource driven (content-less) sites?

support
1

I have decided to build a “content-less” site that will be generated from a snapshot view of a (generic) git repository. Which means there will be potentially thousands of generated pages for the objects: heads, tags, commits all the way to tree entries and actual blobs (taking advantage of chroma to show it all).

(Feel free to tell me it’s a ridiculous use of Hugo, but I find it a great match so far, even kind of enjoy the journey.)

I am currently generating a bunch of JSON files, which I put into assets/ (mostly because they are accompanied by non-resource-like .diff deltas which then get rendered for a comprehensive commit view).

It all works surprisingly well so far with a tiny repo. So this got me wondering where is the limit and how to get closer to it.

I figured early on that I absolutely have to generate index.md’s in the right paths no matter what (not a problem), but I somehow noticed large front matter tends to slow down generation, as if more so than processing the JSONs as a resource from virtually empty pages, hence the decision to keep everything resource-driven.

I wonder if there’s any well-established tips for sites like these (well, of this nature): no front matter, absolutely empty .Content - everything getting taken care of by the templates - and very little change (virtually none) in once-generated resources - additions only, but potentially thousands of (inter-linked) pages.

I have the ability to limit what I generate in terms of JSON, I can generate different format (also with different internal structure, depth, etc.). I am trying to avoid map lookups, sorting or calculations in the templates, mostly doing range over the JSON structure.

But I do not know if it would not have been e.g. better to keep the JSON in page bundles (with publishResources: false) instead of one centralized directory. I also do not know if I should not e.g. first load it all an cache rather than let it crawl the filesystem. Currently, I have no list pages simply because I build such from a JSON list instead, which I feel is going easy on Hugo.

Anyone with experience with anything of this sorts, I’d love to hear any tips early on before I bog myself down in the wrong paradigm. :slight_smile:

Thanks a lot!

2

So, doing the “lots of JSON files in assets” and loading them via transform.Unmarshal is the right approach, the main reason is that Hugo can garbage collect them when they’re not in use anymore. This is not the case for front matter data. JSON files unmarshaled to Go maps can be memory hungry. As to JSON page resources, I don’t see any obvious benefits for your use case.