X-Git-Url: http://git.pjr.cc/?p=nodejs-repoproxy.git;a=blobdiff_plain;f=lib%2Fcache.js;h=9e5e8daf2cb5b630d86f9a91cc25ecf9a1c0ce86;hp=e065a2a160806d587a5ef1833b21d55f36c0ff51;hb=43d51752ca1aafe7212ee3805ba1b33e398c91ce;hpb=68fbf923faf99f86cc36dc5776451f89925df8bd diff --git a/lib/cache.js b/lib/cache.js index e065a2a..9e5e8da 100644 --- a/lib/cache.js +++ b/lib/cache.js @@ -1,6 +1,7 @@ var fs = require("fs"); var http = require("http"); var url = require("url"); +var path = require("path"); function maintainCache() { // TODO i should check that im already running here and exit if i am @@ -15,7 +16,7 @@ exports.startTimer = function() { setInterval(maintainCache, cacheTimer); } -function upstreamRequest(unify, callback) { +function upstreamRequest(unify) { // first do a head request console.log("upsteram as ", unify.requestFor); @@ -30,13 +31,20 @@ function upstreamRequest(unify, callback) { //unify.b.write("would send to '" + xpath + "'"); //unify.b.end(); + // not doing this properly yet... + if(typeof global.repoproxy.downloads[unify.fullFilePath] != undefined && global.repoproxy.downloads[unify.fullFilePath] == 1) { + console.log("request for file thats being downloaded already, doing inline request"); + inlineService(unify); + return; + } + console.log("sending off to '%s'", xpath); var headReq = url.parse(xpath); headReq["method"] = "HEAD"; - getup = http.request(xpath, function(res) { - res.setEncoding("utf8"); + getup = http.request(headReq, function(res) { + //res.setEncoding("utf8"); if(!endData) { console.log("status code is ", typeof res.statusCode); @@ -75,9 +83,14 @@ function upstreamRequest(unify, callback) { endData = true; } else { // this is where it gets ugly + var filesize = res.headers["content-length"]; console.log("do ugly write: ", unify); //unify.b.write(data); - getAndService(unify, xpath); + var metafilename = unify.fullPathDirName + "/.meta."+ path.basename(unify.requestFor) +".filesize"; + var metafile = fs.createWriteStream(metafilename); + metafile.write(filesize); + metafile.end(); + getAndService(unify, xpath, filesize); } break; @@ -96,58 +109,179 @@ function upstreamRequest(unify, callback) { exports.upstreamRequest = upstreamRequest; -function getAndService(unify, xpath) { +function getAndService(unify, xpath, filesize) { - if(global.repoproxy.downloads[unify.fullFilePath] == 1) { - - unify.b.write("trying to service inline"); - unify.b.end(); - } else { - global.repoproxy.downloads[unify.fullFilePath] = 1; + console.log("calling in here with filesize, ", filesize) + unify.b.writeHead(200, {'Content-Length' : filesize}); + - http.get(xpath, function(res) { + global.repoproxy.downloads[unify.fullFilePath] = 1; - var file = fs.createWriteStream(unify.fullFilePath); - - console.log("res: ", res); + + http.get(xpath, function(res) { + + var file = fs.createWriteStream(unify.fullFilePath); + + //console.log("res: ", res); + + //res.setEncoding("utf8"); + + res.on("data", function(data) { + //console.log("chunk"); + file.write(data); + unify.b.write(data); + }); + + res.on("end", function() { + console.log("end..."); + unify.b.end(); + file.end(); + global.repoproxy.downloads[unify.fullFilePath] = 0; + }); + + res.on("error", function(err) { + console.log("res threw error... ", err); + }); + }); +} + +// this is nasty nasty thing that can go horribly wrong in some ways, but currently works... +function inlineService(unify) { + // this method is called when we need to service a file thats being downloaded by something else + var metafilename = unify.fullPathDirName + "/.meta."+ path.basename(unify.requestFor) +".filesize"; + var fsizef = fs.createReadStream(metafilename); + var fsize = ""; + var lastchunk = 0; + fsizef.on("data", function(data) { + fsize += data; + }); + + fsizef.on("end", function() { + var sentSoFar = 0; + unify.b.writeHead(200, {"Content-Length" : fsize }); - //res.setEncoding("utf8"); + // now we go into the file reading loop. + console.log("start of inline services"); + // we loop every 0.5s and do our thing - res.on("data", function(data) { - //console.log("chunk"); - file.write(data); - unify.b.write(data); - }); + function sendPieces() { + // this is going to be so fun i want to play real life frogger in real life traffic... + fs.stat(unify.fullFilePath, function(err, stats) { + if(err == null) { + if(stats["size"] > sentSoFar) { + // if file size changed between last chunk and this chunk, send the chunks + + lastChunk = 0; + // open the file, send the data + var rs = fs.createReadStream(unify.fullFilePath, {start: sentSoFar, end: stats["size"]}); + + rs.on("data", function(thisdata) { + //console.log("inline chunk: ", thisdata.length); + unify.b.write(thisdata); + }); + + rs.on("end", function() { + sentSoFar = stats["size"]; + // every second, we start again + if(sentSoFar != fsize) { + setTimeout(sendPieces, 1000); + } else { + // we're done! + unify.b.end(); + } + }); + } else { + // if file size did not change between last timeout and this one, incremement the chunk counter + // if we reach 60, we had a problem, and so we bomb out + + lastChunk++; + + // we bombed out somehow + if(lastChunk > 60) { + unify.b.end(); + } else { + setTimeout(sendPieces, 1000); + } + } + } else { + console.log("inline service - we're in a very bad place"); + } + }); + + } - res.on("end", function() { - console.log("end..."); - unify.b.end(); - file.end(); - global.repoproxy.downloads[unify.fullFilePath] = 0; - }); - }); - } + setTimeout(sendPieces, 100); + }); } // the service file routine .... PLEASE KILL ME! function serviceFile(unify) { // for now, ignore range. + // however we need to check if a metadata file exists describing the filesize, check if its all correct + // and if not, erase the file (and metafile) and forward the request back to upstream request - // file should already exist, so we just poop it out - var inp = fs.createReadStream(unify.fullFilePath); - inp.setEncoding("utf8"); - inp.on("data", function(data) { - unify.b.write(data); - }); - inp.on("end", function(closed) { - unify.b.end(); + checkFile(unify, function() { + + // file should already exist, so we just poop it out + var inp = fs.createReadStream(unify.fullFilePath); + //inp.setEncoding("utf8"); + inp.on("data", function(data) { + unify.b.write(data); + }); + + inp.on("end", function(closed) { + unify.b.end(); + }); }); } exports.serviceFile = serviceFile; + +function checkFile(unify, callback) { + // in here we do the metadata checks + var metafilename = unify.fullPathDirName + "/.meta."+ path.basename(unify.requestFor) +".filesize"; + + fs.exists(metafilename, function(existence) { + if(existence) { + var fsizef = fs.createReadStream(metafilename); + var fsize = ""; + fsizef.on("data", function(data) { + fsize += data; + }); + + fsizef.on("end", function() { + fs.stat(unify.fullFilePath, function(err, stats) { + var rfsize = stats["size"]; + if(rfsize != fsize.trim()) { + // remove the file and start again + console.log("reported filesizes dont match, '%s', '%s', removing file and starting again", rfsize, stats["size"]); + try { + fs.unlink(metafilename, function(){ + fs.unlink(unify.fullFilePath, function(){ + upstreamRequest(unify); + }) + }); + } catch(e) { + upstreamRequest(unify); + } + } else { + // we're good + unify.b.writeHead(200, {"Content-Length" : unify.fileSize}) + callback(); + } + }); + }); + } else { + console.log("file, '%s' exists but has no filesize meta data, assuming it was put here manually and servicing", unify.fullFilePath); + unify.b.writeHead(200, {"Content-Length" : unify.fileSize}) + callback(); + } + }); +} + function makeCacheDir(path) { console.log("attempting to create... '%s' as '%s'", path.fullPathDirName, path.subPathDirName); @@ -170,7 +304,7 @@ function serviceDirectory(unify) { var res = unify.b; res.write("

Directory listing for " + unify.originalReq + "


");
-	if(unify.fullFilePath != "/") res.write("Parent\n\n");
+	if(unify.originalReq != "/") res.write("Parent\n\n");
 	fs.readdir(unify.fullFilePath, function(err, files) {
 		console.log("doing directory listing on: ", unify.fullFilePath);
 		if(err == null) {