Using the C preprocessor as an HTML templating engine | Hacker News Reader

Using the C preprocessor as an HTML templating engine

github.com

39 pointscedric_h31 comments

IIRC the earliest version of the C++ language, before a C++ compiler existed, was implemented with C and the M4 preprocessor.

https://en.wikipedia.org/wiki/M4_(computer_language)

https://www.gnu.org/software/m4/

woah, that's a really fun tidbit.

Yep, it was called cfront.

https://en.wikipedia.org/wiki/Cfront

cfront was the actual compiler; it was written in c++, not m4

cfront was a C++ => C transpiler

Thank you, I did not know this. I was about to comment and say: Why not go with m4(1)? I have used it for a personal homepage and while it certainly will try to stab you in the back like any old macro processor will, it is more full-fledged that its C counterpart and will also allow to to smugly state that your homepage is POSIX compliant [1]. There are a few guides out there from around 2000 to get you started, but at some point one needs to read the original paper [2] and let your imagination run wild.

[1]: https://pubs.opengroup.org/onlinepubs/9699919799/utilities/m...

[2]: https://maibriz.de/unix/ultrix/etc/m4.pdf

one of the first things i did with html was to attempt to use m4 to simplify my personal home page. not only did that not work, it backfired to an incredible degree, but that's because i was using proprietary vendor m4. gnu m4 has a couple of features that make it significantly less error-prone:

1. built-in macro names that require arguments such as `define' are not "expanded" if you don't follow them with parentheses

2. there's an option to require a magic sigil character to invoke macros, such as % or something; this turns out to have been required in the original gpm and its bell labs clone m6. this helps enormously with complicated macros

however, the fact that the output of your macro is always immediately run as new code (as opposed to m6 and gpm, where this was optional on a per-call basis) makes m4 just ridiculously error-prone. i feel like trauma from debugging m4 macros was a major consideration in the design of the weak-ass c preprocessor

> 1. built-in macro names that require arguments such as `define' are not "expanded" if you don't follow them with parentheses

This one is pure pain. Thankfully, the only implementation I have encountered that does this is Heirloom's [1]. I doubt POSIX will do it, but I wish they would just define this behaviour out of existence for the built-ins as it is a minefield when trying to write prose (or comments, for that matter).

[1]: https://heirloom.sourceforge.net

As for outputs being interpreted iteratively, there is no way around it that I know of. To make invocation more explicit, I have just settled on prefixing macros with an underscore and it has worked well so far as it is uncommon in English text, HTML, and CSS.

Seems like a bad idea, given that the CPP and HTML have a fair number of overlapping tokens (e.g. `#` for both URL fragments and macro beginnings).

Why not use SSI[1]? It's inspired by the CPP's syntax, but is supported directly by most HTTP servers (and is more powerful, to boot).

[1]: https://en.wikipedia.org/wiki/Server_Side_Includes

# is only processed by cpp when it's the first non-whitespace character in a line so that shouldn't be an issue in the real world. But, yeah, SSI.

True. Maybe < and > would have been better examples, although one could argue those are optional :-)

Yes, although if you use #define or C comments or several other stuff then it can potentially cause a problem due to it is not a C syntax and is not expected.

>Why not use SSI[1]?

Because hacking is fun, and it's still fun if it's a bad idea

Thank you for saying this!

Completely forgot about SSI! What a cool feature

I do this for some very small static sites like: [0]

[0] http://rbl.oc9.org/

Very cool, I like the simple mark up.

Quite a long time ago I worked at a company developing games for mobile phones in J2ME. The J2ME runtime had a lot of (different) quirks issues on different phones and the phones where also quite different. We used the C preprocessor a lot to make porting the games for different models easier.

There's also the htp processor that handles the same problem. It has macros (equivalent to #define) file inclusion (equivalent to #include) and conditionals (equivalent to #ifxxx) but is better suited for HTML than the C preprocessor. https://sourceforge.net/projects/htp/

I was considering doing something like this just the other day. It had a gnuplot script and it seemed to me that it would be so much more versatile if I could use #ifdefine and #include in it.

I ended up using a Perl preprocessor, but it seems like cpp would work ok-ish for certain use cases that don’t require, say, strict tabs like Python among a few other things mentioned on the cpp manual.

Check out https://github.com/logological/gpp it's what I use when I feel like I need a preprocessor. It's more flexible in syntax and operation than cpp.

I've used it in the past for wrangling the definitions of a multitude of SQL triggers I was using to generate a real-time materialized view. The fact that you can modify its syntax is very helpful for making it work with non-C syntaxes (like SQL or even LaTeX).

It even has an HTML mode built-in, which you can customize to your needs, but looks like:

    <#define x|y>
    <#macro arg|...>

which works better for certain WYSIWYG editors. You could modify the syntax to be

    <!--#define x y-->
    <!--#macro(arg, ...)-->

Thanks for sharing! Looks just like what I need :-)

Why not go all out? Use -DUSERNAME=$USERNAME on the command line to pass in data? Define macros with parameters to affect rendering in complicated ways. Call cpp for every request to generate the html. Could be fun!

I've gotten a series of deranged preprocessor hacks from friends since sharing that repo, the possibilities are ... terrifyingly limitless.

https://news.ycombinator.com/item?id=39863492

nice!

Or there's the opposite choice of using PHP as a C preprocessors. Not sure which is more frightening.

They're both pretty chill-inducing.

i've used a linux setup for wireless access points that used the unix shell as its html templating engine. (openwrt 20 years ago, maybe?) the web server would run a shell script and send its output to the browser, cgi-style

Reply on news.ycombinator.com