Roadmap

[bebraw]

Snipped from #129.

I'm currently refactoring it a bit (simpler configuration) and hope to get the work done very soon so I can start using the generator in places. I've actually pushed a lot of weight from the core during this process. This has given a nice degree of flexibility that I like. I'm also adding some tests in place so it's easier to contribute if someone wants to.

There's no official road map but I have at least the following concerns in mind:

• Performance - Now it generates initial bundle and then uses that to generate each page. I believe it would be better to parallelize this (worker process, nothing special) and generate a bundle per instance (less to parse). This won't help with Travis or single threaded platforms but it would likely help in my personal usage.
• Routing - The current approach doesn't give you that SPA feeling. I think a little different approach may be required. Possibly a router could be implemented on top of a generated site although I can also consider generating JSON etc. to load PWA style.
• Replace React with rapscallion - I gave this a go earlier but I hit some issue with rapscallion. When it's mature enough, moving to it during generation might be a good way to go.
• Incremental generation - Instead of generating the whole site always, it would be good to look into caching at least a part of it. Maybe things like cache-loader can help here. Caching is a difficult problem, though, and I may have to invent new ways to deal with it (compose a site out of fragments that are then invalidated based on change).

Especially the last items might not happen as I don't have to maintain big sites to warrant that style of work. That said, I have to push the performance somehow as a Travis build against the webpack site is far too slow. A part of that might be just a matter of configuration.

[sukrosono]

The last item really appealing, yet we don't have better approach.

generating the whole site always

What about we prevent creating new file (only html?) , and look existing file. When we found we compare the date, take the process the new one. Also introduce new flag always_generate , that prevent the default behavior (described above).

It's just theoretical approach, i am not using antwar yet. I on my way learning react and found antwar when examine webpack documentation (webpack.js.org).

[bebraw]

@brutalcrozt Caching might be a red herring. Since I wrote the issue, Antwar has become significantly faster. It's due to architectural improvements, webpack, and Node speedups. I can generate a site with about 300 pages in 5 seconds on my computer. Speed is not an issue. 👍

You can try it yourself by cloning this site.

[bebraw]

I've worked on performance a lot lately. The generator runs on webpack 4 now and I also found a few ways to improve the configuration further.