#!/usr/bin/env retro ================================================================ ____ _ _ __ / ___|__ _ ___| | _____| |_ __ __/ /_ | | / _` / __| |/ / _ \ __| \ \ / / '_ \ | |__| (_| \__ \ < __/ |_ \ V /| (_) | \____\__,_|___/_|\_\___|\__| \_/ \___/ ================================================================ Casket is a small HTTP/1.1 server written in RetroForth. Some notes on this: - runs under inetd - use w/stunnel if you need HTTPS - I recommend also using timelimit(1) with this Changes from v5: - removed support for running .fsp files - now using `file:read/bytes` and `file:write/bytes` - support for HEAD requests - reject POST, PUT, DELETE requests - various refactorings ================================================================ First, some configuration options. Since this will run under inetd there's no need to specify the port. But the path to the files to serve is needed, so define it here. ~~~ '/home/crc/wip-test 'WEBROOT s:const ~~~ The server supports virtual servers. For this, create a separate directory for each under the `WEBROOT`. E.g., I use a setup like /www/retroforth.org /www/retroforth.org:443 /www/ilo.retroforth.org Note the port number: if using a port other than 80 this is needed. I recommend just using a symlink. ================================================================ Next, I need to handle the incoming requests. HTTP allows for a large number of header fields, but I really only care about two: GET and Host. An incoming request will look like: GET / HTTP/1.1 Host: retroforth.org With the lines ending in a CR,LF sequence. I need to allocate space for the data I care about. There are three items: - The `Requested` file - The desired virtual `Host` - The query string (if any) ~~~ 'Requested d:create #8193 allot 'Host d:create #1025 allot 'Method var 'Query var ~~~ The header processor will read each item and store the `Host` and `Requested` file. Everything else is ignored. I implement `eot?` to decide if a line (or field) indicator has been reached. This is used by `s:get` to decide when the input should stop. `s:get` records the characters read into the specified buffer. And finally, `read-request` reads the input. ~~~ 'Done var :eot? (c-f) [ ASCII:CR eq? ] [ ASCII:LF eq? ] [ ASCII:SPACE eq? ] tri or or ; :s:get (a-) buffer:set [ c:get &buffer:add &eot? bi ] until buffer:get drop ; :req? (s-f) here s:to-upper swap s:eq? ; :read-request (-) [ here s:get 'GET req? [ &Requested s:get &Done v:inc #1 !Method ] if 'HEAD req? [ &Requested s:get &Done v:inc #2 !Method ] if (unsupported) 'PUT req? [ &Requested s:get &Done v:inc #0 !Method ] if 'POST req? [ &Requested s:get &Done v:inc #0 !Method ] if 'DELETE req? [ &Requested s:get &Done v:inc #0 !Method ] if 'HOST: req? [ &Host s:get &Done v:inc ] if @Done #2 eq? ] until ; ~~~ ================================================================ Next is reading in the desired file. An initial request may be just a **/**. In this case, Casket will replace the `Requested` filename with **/index.html**. In the odd case that a file is requested without a leading **/**, I have a word to add this. And then a word that constructs a filename. This also has a word `check-for-params` that is used to separate the requested file from any query string that may be present. ~~~ :map-/-to-index (-) &Requested '/ s:eq? [ '/index.html &Requested s:copy ] if ; :ensure-leading-/ (-) @Requested $/ -eq? [ '/ &Requested s:append s:keep &Requested s:copy ] if ; :s:terminate (s-) #0 swap store ; :check-for-params (-) &Requested $? s:contains/char? [ &Requested $? s:split/char drop [ n:inc !Query ] &s:terminate bi ] if ; :prepare check-for-params map-/-to-index ensure-leading-/ ; :filename (-s) prepare &Requested &Host WEBROOT '%s/%s%s s:format dup '%s/index.html s:format file:exists? [ '%s/index.html s:format ] if ; ~~~ Next, I need to determine the file type. I'll do this by taking a look at the extension, and mapping this to a MIME type. ~~~ :filename-w/o-path WEBROOT s:length &Host s:length + filename + [ $/ s:index/char ] sip + ; :get-mime-type (-s) filename-w/o-path [ $. s:index/char ] sip + (textual_files) '.md [ 'text/markdown ] s:case '.htm [ 'text/html ] s:case '.html [ 'text/html ] s:case '.css [ 'text/css ] s:case (image_files) '.png [ 'image/png ] s:case '.jpg [ 'image/jpeg ] s:case '.jpeg [ 'image/jpeg ] s:case '.gif [ 'image/gif ] s:case '.bmp [ 'image/bmp ] s:case (application_files) '.pdf [ 'application/pdf ] s:case '.gz [ 'application/gzip ] s:case '.zip [ 'application/zip ] s:case '.json [ 'application/json ] s:case '.js [ 'application/x-javascript ] s:case '.xml [ 'application/xml ] s:case '.tar.gz [ 'application/gzip ] s:case '.retro.html [ 'text/html ] s:case drop 'text/plain ; ~~~ ================================================================ Using these, I can construct a word to read in the file and send it to the client. Reading files is now a bit more involved, since images and other formats have binary data. `transfer` performs the actual process of reading the requested file and sending it to the client. This makes use of the `file:read/bytes` and `file:write/bytes` to do the actual i/o. These are used as they are faster than reading & writing on a per-byte basis. Adjust the `FileBuffer` size (and matching size constant) as desired. The requsted file will be read in chunks of this size, so matching the splits better for te files you are using can aid in performance. ~~~ {{ 'FID var 'FOUT var 'ChunkSize var 'FileBuffer d:create #16385 allot :open-files (-) filename file:R file:open !FID '/dev/stdout file:W file:open !FOUT ; :prepare (-nn) #16384 !ChunkSize @FID file:size @ChunkSize n:divmod ; :read (-) &FileBuffer @ChunkSize @FID file:read/bytes drop ; :write (-) &FileBuffer @ChunkSize @FOUT file:write/bytes drop ; :chunk (-) read write ; :cleanup (-) @FID file:close @FOUT file:close ; ---reveal--- :transfer (-) open-files prepare &chunk times !ChunkSize chunk cleanup ; }} ~~~ ~~~ :eol (-) ASCII:CR c:put ASCII:LF c:put ; :send-file (-) get-mime-type 'Content-type:_%s s:format s:put eol eol @Method #1 eq? &transfer if ; ~~~ In the above, `eol` will send an end of line sequence. ================================================================ The last support word is a handler for 404 errors. This will send the 404 status code and a human readable error message. ~~~ :404 'HTTP/1.1_404_OK s:put eol 'Content-type:_text/html s:put eol eol '