UNPKG

link-rdflib

Version:

an RDF library for node.js, patched for speed.

1,556 lines (1,345 loc) 54.5 kB
/* global $SolidTestEnvironment */ /** * * Project: rdflib.js * * @file: fetcher.js * * Description: contains functions for requesting/fetching/retracting * This implements quite a lot of the web architecture. * A fetcher is bound to a specific quad store, into which * it loads stuff and into which it writes its metadata * @@ The metadata could be optionally a separate graph * * - implements semantics of HTTP headers, Internet Content Types * - selects parsers for rdf/xml, n3, rdfa, grddl * * TO do: * - Implement a runtime registry for parsers and serializers * - */ /** * Things to test: callbacks on request, refresh, retract * loading from HTTP, HTTPS, FTP, FILE, others? * To do: * Firing up a mail client for mid: (message:) URLs */ const IndexedFormula = require('./store') const log = require('./log') const N3Parser = require('./n3parser') const NamedNode = require('./named-node') const Namespace = require('./namespace') const rdfParse = require('./parse') const parseRDFaDOM = require('./rdfaparser').parseRDFaDOM const RDFParser = require('./rdfxmlparser') const Term = require('./term') const Uri = require('./uri') const Util = require('./util') const serialize = require('./serialize') // This is a special fetch withich does OIDC auth, catching 401 errors const auth = require('solid-auth-client') const fetch = auth ? auth.fetch : (a, b) => window.fetch(a, b) const Parsable = { 'text/n3': true, 'text/turtle': true, 'application/rdf+xml': true, 'application/xhtml+xml': true, 'text/html': true, 'application/ld+json': true } // This is a minimal set to allow the use of damaged servers if necessary const CONTENT_TYPE_BY_EXT = { 'rdf': 'application/rdf+xml', 'owl': 'application/rdf+xml', 'n3': 'text/n3', 'ttl': 'text/turtle', 'nt': 'text/n3', 'acl': 'text/n3', 'html': 'text/html', 'xml': 'text/xml' } // Convenience namespaces needed in this module. // These are deliberately not exported as the user application should // make its own list and not rely on the prefixes used here, // and not be tempted to add to them, and them clash with those of another // application. const ns = { link: Namespace('http://www.w3.org/2007/ont/link#'), http: Namespace('http://www.w3.org/2007/ont/http#'), httph: Namespace('http://www.w3.org/2007/ont/httph#'), // headers rdf: Namespace('http://www.w3.org/1999/02/22-rdf-syntax-ns#'), rdfs: Namespace('http://www.w3.org/2000/01/rdf-schema#'), dc: Namespace('http://purl.org/dc/elements/1.1/'), ldp: Namespace('http://www.w3.org/ns/ldp#') } class Handler { constructor (response, dom) { this.response = response this.dom = dom } } class RDFXMLHandler extends Handler { static toString () { return 'RDFXMLHandler' } static register (fetcher) { fetcher.mediatypes['application/rdf+xml'] = { 'q': 0.9 } } parse (fetcher, responseText, options, response) { let kb = fetcher.store if (!this.dom) { this.dom = Util.parseXML(responseText) } let root = this.dom.documentElement if (root.nodeName === 'parsererror') { // Mozilla only See issue/issue110 // have to fail the request return fetcher.failFetch(options, 'Badly formed XML in ' + options.resource.uri, 'parse_error') } let parser = new RDFParser(kb) try { parser.parse(this.dom, options.original.uri, options.original, response) } catch (err) { return fetcher.failFetch(options, 'Syntax error parsing RDF/XML! ' + err, 'parse_error') } if (!options.noMeta) { kb.add(options.original, ns.rdf('type'), ns.link('RDFDocument'), fetcher.appNode) } return fetcher.doneFetch(options, this.response) } } RDFXMLHandler.pattern = new RegExp('application/rdf\\+xml') class XHTMLHandler extends Handler { static toString () { return 'XHTMLHandler' } static register (fetcher) { fetcher.mediatypes['application/xhtml+xml'] = {} } parse (fetcher, responseText, options, response) { let relation, reverse if (!this.dom) { this.dom = Util.parseXML(responseText) } let kb = fetcher.store // dc:title let title = this.dom.getElementsByTagName('title') if (title.length > 0) { kb.add(options.resource, ns.dc('title'), kb.literal(title[0].textContent), options.resource) // log.info("Inferring title of " + xhr.resource) } // link rel let links = this.dom.getElementsByTagName('link') for (let x = links.length - 1; x >= 0; x--) { // @@ rev relation = links[x].getAttribute('rel') reverse = false if (!relation) { relation = links[x].getAttribute('rev') reverse = true } if (relation) { fetcher.linkData(options.original, relation, links[x].getAttribute('href'), options.resource, reverse) } } // Data Islands let scripts = this.dom.getElementsByTagName('script') for (let i = 0; i < scripts.length; i++) { let contentType = scripts[i].getAttribute('type') if (Parsable[contentType]) { rdfParse(scripts[i].textContent, kb, options.original.uri, contentType) rdfParse(scripts[i].textContent, kb, options.original.uri, contentType) } } if (!options.noMeta) { kb.add(options.resource, ns.rdf('type'), ns.link('WebPage'), fetcher.appNode) } if (!options.noRDFa && parseRDFaDOM) { // enable by default try { parseRDFaDOM(this.dom, kb, options.original.uri) } catch (err) { let msg = 'Error trying to parse ' + options.resource + ' as RDFa:\n' + err + ':\n' + err.stack return fetcher.failFetch(options, msg, 'parse_error') } } return fetcher.doneFetch(options, this.response) } } XHTMLHandler.pattern = new RegExp('application/xhtml') class XMLHandler extends Handler { static toString () { return 'XMLHandler' } static register (fetcher) { fetcher.mediatypes['text/xml'] = { 'q': 0.5 } fetcher.mediatypes['application/xml'] = { 'q': 0.5 } } parse (fetcher, responseText, options, response) { let dom = Util.parseXML(responseText) // XML Semantics defined by root element namespace // figure out the root element for (let c = 0; c < dom.childNodes.length; c++) { // is this node an element? if (dom.childNodes[c].nodeType === 1) { // We've found the first element, it's the root let ns = dom.childNodes[c].namespaceURI // Is it RDF/XML? if (ns && ns === ns['rdf']) { fetcher.addStatus(options.req, 'Has XML root element in the RDF namespace, so assume RDF/XML.') let rdfHandler = new RDFXMLHandler(this.response, dom) return rdfHandler.parse(fetcher, responseText, options, response) } break } } // Or it could be XHTML? // Maybe it has an XHTML DOCTYPE? if (dom.doctype) { // log.info("We found a DOCTYPE in " + xhr.resource) if (dom.doctype.name === 'html' && dom.doctype.publicId.match(/^-\/\/W3C\/\/DTD XHTML/) && dom.doctype.systemId.match(/http:\/\/www.w3.org\/TR\/xhtml/)) { fetcher.addStatus(options.req, 'Has XHTML DOCTYPE. Switching to XHTML Handler.\n') let xhtmlHandler = new XHTMLHandler(this.response, dom) return xhtmlHandler.parse(fetcher, responseText, options, response) } } // Or what about an XHTML namespace? let html = dom.getElementsByTagName('html')[0] if (html) { let xmlns = html.getAttribute('xmlns') if (xmlns && xmlns.match(/^http:\/\/www.w3.org\/1999\/xhtml/)) { fetcher.addStatus(options.req, 'Has a default namespace for ' + 'XHTML. Switching to XHTMLHandler.\n') let xhtmlHandler = new XHTMLHandler(this.response, dom) return xhtmlHandler.parse(fetcher, responseText, options, response) } } // At this point we should check the namespace document (cache it!) and // look for a GRDDL transform // @@ Get namespace document <n>, parse it, look for <n> grddl:namespaceTransform ?y // Apply ?y to dom // We give up. What dialect is this? return fetcher.failFetch(options, 'Unsupported dialect of XML: not RDF or XHTML namespace, etc.\n' + responseText.slice(0, 80), 901) } } XMLHandler.pattern = new RegExp('(text|application)/(.*)xml') class HTMLHandler extends Handler { static toString () { return 'HTMLHandler' } static register (fetcher) { fetcher.mediatypes['text/html'] = { 'q': 0.9 } } parse (fetcher, responseText, options, response) { let kb = fetcher.store // We only handle XHTML so we have to figure out if this is XML // log.info("Sniffing HTML " + xhr.resource + " for XHTML.") if (responseText.match(/\s*<\?xml\s+version\s*=[^<>]+\?>/)) { fetcher.addStatus(options.req, "Has an XML declaration. We'll assume " + "it's XHTML as the content-type was text/html.\n") let xhtmlHandler = new XHTMLHandler(this.response) return xhtmlHandler.parse(fetcher, responseText, options, response) } // DOCTYPE // There is probably a smarter way to do this if (responseText.match(/.*<!DOCTYPE\s+html[^<]+-\/\/W3C\/\/DTD XHTML[^<]+http:\/\/www.w3.org\/TR\/xhtml[^<]+>/)) { fetcher.addStatus(options.req, 'Has XHTML DOCTYPE. Switching to XHTMLHandler.\n') let xhtmlHandler = new XHTMLHandler(this.response) return xhtmlHandler.parse(fetcher, responseText, options, response) } // xmlns if (responseText.match(/[^(<html)]*<html\s+[^<]*xmlns=['"]http:\/\/www.w3.org\/1999\/xhtml["'][^<]*>/)) { fetcher.addStatus(options.req, 'Has default namespace for XHTML, so switching to XHTMLHandler.\n') let xhtmlHandler = new XHTMLHandler(this.response) return xhtmlHandler.parse(fetcher, responseText, options, response) } // dc:title // no need to escape '/' here let titleMatch = (new RegExp('<title>([\\s\\S]+?)</title>', 'im')).exec(responseText) if (titleMatch) { kb.add(options.resource, ns.dc('title'), kb.literal(titleMatch[1]), options.resource) // think about xml:lang later } kb.add(options.resource, ns.rdf('type'), ns.link('WebPage'), fetcher.appNode) fetcher.addStatus(options.req, 'non-XML HTML document, not parsed for data.') return fetcher.doneFetch(options, this.response) } } HTMLHandler.pattern = new RegExp('text/html') class TextHandler extends Handler { static toString () { return 'TextHandler' } static register (fetcher) { fetcher.mediatypes['text/plain'] = { 'q': 0.5 } } parse (fetcher, responseText, options, response) { // We only speak dialects of XML right now. Is this XML? // Look for an XML declaration if (responseText.match(/\s*<\?xml\s+version\s*=[^<>]+\?>/)) { fetcher.addStatus(options.req, 'Warning: ' + options.resource + " has an XML declaration. We'll assume " + "it's XML but its content-type wasn't XML.\n") let xmlHandler = new XMLHandler(this.response) return xmlHandler.parse(fetcher, responseText, options, response) } // Look for an XML declaration if (responseText.slice(0, 500).match(/xmlns:/)) { fetcher.addStatus(options.req, "May have an XML namespace. We'll assume " + "it's XML but its content-type wasn't XML.\n") let xmlHandler = new XMLHandler(this.response) return xmlHandler.parse(fetcher, responseText, options, response) } // We give up finding semantics - this is not an error, just no data fetcher.addStatus(options.req, 'Plain text document, no known RDF semantics.') return fetcher.doneFetch(options, this.response) } } TextHandler.pattern = new RegExp('text/plain') class N3Handler extends Handler { static toString () { return 'N3Handler' } static register (fetcher) { fetcher.mediatypes['text/n3'] = { 'q': '1.0' } // as per 2008 spec /* fetcher.mediatypes['application/x-turtle'] = { 'q': 1.0 } // pre 2008 */ fetcher.mediatypes['text/turtle'] = { 'q': 1.0 } // post 2008 } parse (fetcher, responseText, options, response) { // Parse the text of this N3 file let kb = fetcher.store let p = N3Parser(kb, kb, options.original.uri, options.original.uri, null, null, '', null) // p.loadBuf(xhr.responseText) try { p.loadBuf(responseText) } catch (err) { let msg = 'Error trying to parse ' + options.resource + ' as Notation3:\n' + err // not err.stack -- irrelevant return fetcher.failFetch(options, msg, 'parse_error', response) } fetcher.addStatus(options.req, 'N3 parsed: ' + p.statementCount + ' triples in ' + p.lines + ' lines.') fetcher.store.add(options.original, ns.rdf('type'), ns.link('RDFDocument'), fetcher.appNode) return fetcher.doneFetch(options, this.response) } } N3Handler.pattern = new RegExp('(application|text)/(x-)?(rdf\\+)?(n3|turtle)') const HANDLERS = { RDFXMLHandler, XHTMLHandler, XMLHandler, HTMLHandler, TextHandler, N3Handler } /** Fetcher * * The Fetcher object is a helper object for a quadstore * which turns it from an offline store to an online store. * The fetcher deals with loading data files rom the web, * figuring how to parse them. It will also refresh, remove, the data * and put back the fata to the web. */ class Fetcher { /** * @constructor */ constructor (store, options = {}) { this.store = store || new IndexedFormula() this.timeout = options.timeout || 30000 this._fetch = options.fetch || fetch if (!this._fetch) { throw new Error('No _fetch function availble for Fetcher') } this.appNode = Object.hasOwnProperty.call(options, 'graph') || this.store.bnode() // Denoting this session this.store.fetcher = this // Bi-linked this.requested = {} // this.requested[uri] states: // undefined no record of web access or records reset // true has been requested, fetch in progress // 'done' received, Ok // 401 Not logged in // 403 HTTP status unauthorized // 404 Resource does not exist. Can be created etc. // 'redirected' In attempt to counter CORS problems retried. // 'parse_error' Parse error // 'unsupported_protocol' URI is not a protocol Fetcher can deal with // other strings mean various other errors. // this.timeouts = {} // list of timeouts associated with a requested URL this.redirectedTo = {} // When 'redirected' this.fetchQueue = {} this.fetchCallbacks = {} // fetchCallbacks[uri].push(callback) this.nonexistent = {} // keep track of explicit 404s -> we can overwrite etc this.lookedUp = {} this.handlers = [] this.mediatypes = { 'image/*': { 'q': 0.9 }, '*/*': { 'q': 0.1 } // Must allow access to random content } // Util.callbackify(this, ['request', 'recv', 'headers', 'load', 'fail', // 'refresh', 'retract', 'done']) // In switching to fetch(), 'recv', 'headers' and 'load' do not make sense Util.callbackify(this, ['request', 'fail', 'refresh', 'retract', 'done']) Object.keys(options.handlers || HANDLERS).map(key => this.addHandler(HANDLERS[key])) } static crossSiteProxy (uri) { if (Fetcher.crossSiteProxyTemplate) { return Fetcher.crossSiteProxyTemplate .replace('{uri}', encodeURIComponent(uri)) } else { return undefined } } /** * @param uri {string} * * @returns {string} */ static offlineOverride (uri) { // Map the URI to a localhost proxy if we are running on localhost // This is used for working offline, e.g. on planes. // Is the script itself is running in localhost, then access all // data in a localhost mirror. // Do not remove without checking with TimBL let requestedURI = uri var UI if (typeof window !== 'undefined' && window.panes && (UI = window.panes.UI) && UI.preferences && UI.preferences.get('offlineModeUsingLocalhost')) { if (requestedURI.slice(0, 7) === 'http://' && requestedURI.slice(7, 17) !== 'localhost/') { requestedURI = 'http://localhost/' + requestedURI.slice(7) log.warn('Localhost kludge for offline use: actually getting <' + requestedURI + '>') } else { // log.warn("Localhost kludge NOT USED <" + requestedURI + ">") } } else { // log.warn("Localhost kludge OFF offline use: actually getting <" + // requestedURI + ">") } return requestedURI } static proxyIfNecessary (uri) { var UI if (typeof window !== 'undefined' && window.panes && (UI = window.panes.UI) && UI.isExtension) { return uri } // Extension does not need proxy if (typeof $SolidTestEnvironment !== 'undefined' && $SolidTestEnvironment.localSiteMap) { // nested dictionaries of URI parts from origin down let hostpath = uri.split('/').slice(2) // the bit after the // const lookup = (parts, index) => { let z = index[parts.shift()] if (!z) { return null } if (typeof z === 'string') { return z + parts.join('/') } if (!parts) { return null } return lookup(parts, z) } const y = lookup(hostpath, $SolidTestEnvironment.localSiteMap) if (y) { return y } } // browser does 2014 on as https browser script not trusted // If the web app origin is https: then the mixed content rules // prevent it loading insecure http: stuff so we need proxy. if (Fetcher.crossSiteProxyTemplate && typeof document !== 'undefined' && document.location && ('' + document.location).slice(0, 6) === 'https:' && // origin is secure uri.slice(0, 5) === 'http:') { // requested data is not return Fetcher.crossSiteProxyTemplate .replace('{uri}', encodeURIComponent(uri)) } return uri } /** * Tests whether the uri's protocol is supported by the Fetcher. * * @param uri {string} * * @returns {boolean} */ static unsupportedProtocol (uri) { let pcol = Uri.protocol(uri) return (pcol === 'tel' || pcol === 'mailto' || pcol === 'urn') } /** * @param requestedURI {string} * @param options {Object} * * @returns {boolean} */ static withCredentials (requestedURI, options = {}) { // 2014 problem: // XMLHttpRequest cannot load http://www.w3.org/People/Berners-Lee/card. // A wildcard '*' cannot be used in the 'Access-Control-Allow-Origin' // header when the credentials flag is true. // @ Many ontology files under http: and need CORS wildcard -> // can't have withCredentials // @@ Kludge -- need for webid which typically is served from https let withCredentials = requestedURI.startsWith('https:') if (options.withCredentials !== undefined) { withCredentials = options.withCredentials } return withCredentials } /** * Promise-based load function * * Loads a web resource or resources into the store. * * A resource may be given as NamedNode object, or as a plain URI. * an arrsy of resources will be given, in which they will be fetched in parallel. * By default, the HTTP headers are recorded also, in the same store, in a separate graph. * This allows code like editable() for example to test things about the resource. * * @param uri {Array<NamedNode>|Array<string>|NamedNode|string} * * @param [options={}] {Object} * * @param [options.fetch] {Function} * * @param [options.referringTerm] {NamedNode} Referring term, the resource which * referred to this (for tracking bad links) * * @param [options.contentType] {string} Provided content type (for writes) * * @param [options.forceContentType] {string} Override the incoming header to * force the data to be treated as this content-type (for reads) * * @param [options.force] {boolean} Load the data even if loaded before. * Also sets the `Cache-Control:` header to `no-cache` * * @param [options.baseURI=docuri] {Node|string} Original uri to preserve * through proxying etc (`xhr.original`). * * @param [options.proxyUsed] {boolean} Whether this request is a retry via * a proxy (generally done from an error handler) * * @param [options.withCredentials] {boolean} flag for XHR/CORS etc * * @param [options.clearPreviousData] {boolean} Before we parse new data, * clear old, but only on status 200 responses * * @param [options.noMeta] {boolean} Prevents the addition of various metadata * triples (about the fetch request) to the store * * @param [options.noRDFa] {boolean} * * @returns {Promise<Result>} */ load (uri, options = {}) { if (uri instanceof Array) { return Promise.all( uri.map(x => { return this.load(x, Object.assign({}, options)) }) ) } let docuri = uri.uri || uri docuri = docuri.split('#')[0] options = this.initFetchOptions(docuri, options) return this.pendingFetchPromise(docuri, options.baseURI, options) } /** * @param uri {string} * @param originalUri {string} * @param options {Object} * @returns {Promise<Result>} */ pendingFetchPromise (uri, originalUri, options) { let pendingPromise // Check to see if some request is already dealing with this uri if (!options.force && this.fetchQueue[originalUri]) { pendingPromise = this.fetchQueue[originalUri] } else { pendingPromise = Promise .race([ this.setRequestTimeout(uri, options), this.fetchUri(uri, options) ]) this.fetchQueue[originalUri] = pendingPromise // Clean up the queued promise after a time, if it's resolved this.cleanupFetchRequest(originalUri, options, this.timeout) } return pendingPromise.then(x => { if (uri in this.timeouts) { this.timeouts[uri].forEach(clearTimeout) delete this.timeouts[uri] } return x }) } cleanupFetchRequest (originalUri, options, timeout) { this.timeouts[originalUri] = (this.timeouts[originalUri] || []).concat(setTimeout(() => { if (!this.isPending(originalUri)) { delete this.fetchQueue[originalUri] } }, timeout)) } /** * @param uri {string} * @param options {Object} * * @returns {Object} */ initFetchOptions (uri, options) { let kb = this.store let isGet = !options.method || options.method.toUpperCase() === 'GET' if (!isGet) { options.force = true } options.resource = kb.sym(uri) // This might be proxified options.baseURI = options.baseURI || uri // Preserve though proxying etc options.original = kb.sym(options.baseURI) options.req = kb.bnode() options.headers = options.headers || {} if (options.contentType) { options.headers['content-type'] = options.contentType } if (options.force) { options.cache = 'no-cache' } let acceptString = this.acceptString() options.headers['accept'] = acceptString let requestedURI = Fetcher.offlineOverride(uri) options.requestedURI = requestedURI if (Fetcher.withCredentials(requestedURI, options)) { options.credentials = 'include' } let actualProxyURI = Fetcher.proxyIfNecessary(requestedURI) if (requestedURI !== actualProxyURI) { options.proxyUsed = true } options.actualProxyURI = actualProxyURI return options } /** * (The promise chain ends in either a `failFetch()` or a `doneFetch()`) * * @param docuri {string} * @param options {Object} * * @returns {Promise<Object>} fetch() result or an { error, status } object */ fetchUri (docuri, options) { if (!docuri) { return Promise.reject(new Error('Cannot fetch an empty uri')) } if (Fetcher.unsupportedProtocol(docuri)) { return this.failFetch(options, 'fetcher: Unsupported protocol', 'unsupported_protocol') } let state = this.getState(docuri) if (!options.force) { if (state === 'fetched') { // URI already fetched and added to store return Promise.resolve( this.doneFetch(options, {status: 200, ok: true, statusText: 'Already loaded into quadstore.'}) ) } if (state === 'failed' && this.requested[docuri] === 404) { // Remember nonexistence let message = 'Previously failed: ' + this.requested[docuri] let dummyResponse = { url: docuri, status: this.requested[docuri], statusText: message, responseText: message, headers: {}, // Headers() ??? ok: false, body: null, bodyUsed: false, size: 0, timeout: 0 } return this.failFetch(options, message, this.requested[docuri], dummyResponse) } } else { // options.force == true delete this.nonexistent[docuri] } this.fireCallbacks('request', [docuri]) this.requested[docuri] = true // mark this uri as 'requested' if (!options.noMeta) { this.saveRequestMetadata(docuri, options) } let { actualProxyURI } = options return this._fetch(actualProxyURI, options) .then(response => this.handleResponse(response, docuri, options), error => { let dummyResponse = { url: actualProxyURI, status: 999, // @@ what number/string should fetch failures report? statusText: (error.name || 'network failure') + ': ' + (error.errno || error.code || error.type), responseText: error.message, headers: {}, // Headers() ??? ok: false, body: null, bodyUsed: false, size: 0, timeout: 0 } console.log('Fetcher: <' + actualProxyURI + '> Non-HTTP fetch error: ' + error) return this.failFetch(options, 'fetch failed: ' + error, 999, dummyResponse) // Fake status code: fetch exception // handleError expects a response so we fake some important bits. /* this.handleError(, docuri, options) */ } ) } /** * Asks for a doc to be loaded if necessary then calls back * * Calling methods: * nowOrWhenFetched (uri, userCallback) * nowOrWhenFetched (uri, options, userCallback) * nowOrWhenFetched (uri, referringTerm, userCallback, options) <-- old * nowOrWhenFetched (uri, referringTerm, userCallback) <-- old * * Options include: * referringTerm The document in which this link was found. * this is valuable when finding the source of bad URIs * force boolean. Never mind whether you have tried before, * load this from scratch. * forceContentType Override the incoming header to force the data to be * treated as this content-type. * * Callback function takes: * * ok True if the fetch worked, and got a 200 response. * False if any error happened * * errmessage Text error message if not OK. * * response The fetch Response object (was: XHR) if there was was one * includes response.status as the HTTP status if any. */ nowOrWhenFetched (uri, p2, userCallback, options = {}) { uri = uri.uri || uri // allow symbol object or string to be passed if (typeof p2 === 'function') { // nowOrWhenFetched (uri, userCallback) userCallback = p2 } else if (typeof p2 === 'undefined') { // original calling signature // referringTerm = undefined } else if (p2 instanceof NamedNode) { // referringTerm = p2 options.referringTerm = p2 } else { // nowOrWhenFetched (uri, options, userCallback) options = p2 } this.load(uri, options) .then(fetchResponse => { if (userCallback) { if (fetchResponse) { if (fetchResponse.ok) { userCallback(true, 'OK', fetchResponse) } else { console.log('@@@ fetcher.js Should not take this path !!!!!!!!!!!!') let oops = 'HTTP error: Status ' + fetchResponse.status + ' (' + fetchResponse.statusText + ')' if (fetchResponse.responseText) { oops += ' ' + fetchResponse.responseText // not in 404, dns error, nock failure } console.log(oops + ' fetching ' + uri) userCallback(false, oops, fetchResponse) } } else { let oops = ('@@ nowOrWhenFetched: no response object!') console.log(oops) userCallback(false, oops) } } }, function (err) { var message = err.message || err.statusText message = 'Failed to load <' + uri + '> ' + message console.log(message) if (err.response && err.response.status) { message += ' status: ' + err.response.status } userCallback(false, message, err.response) }) } /** * Records a status message (as a literal node) by appending it to the * request's metadata status collection. * * @param req {BlankNode} * @param statusMessage {string} */ addStatus (req, statusMessage) { // <Debug about="parsePerformance"> let now = new Date() statusMessage = '[' + now.getHours() + ':' + now.getMinutes() + ':' + now.getSeconds() + '.' + now.getMilliseconds() + '] ' + statusMessage // </Debug> let kb = this.store let statusNode = kb.the(req, ns.link('status')) if (statusNode && statusNode.append) { statusNode.append(kb.literal(statusMessage)) } else { log.warn('web.js: No list to add to: ' + statusNode + ',' + statusMessage) } } /** * Records errors in the system on failure: * * - Adds an entry to the request status collection * - Adds an error triple with the fail message to the metadata * - Fires the 'fail' callback * - Rejects with an error result object, which has a response object if any * * @param options {Object} * @param errorMessage {string} * @param statusCode {number} * @param response {Response} // when an fetch() error * * @returns {Promise<Object>} */ failFetch (options, errorMessage, statusCode, response) { this.addStatus(options.req, errorMessage) if (!options.noMeta) { this.store.add(options.original, ns.link('error'), this.store.literal(errorMessage)) } let meth = (options.method || 'GET').toUpperCase() let isGet = meth === 'GET' || meth === 'HEAD' if (isGet) { // only cache the status code on GET or HEAD if (!options.resource.sameTerm(options.original)) { console.log('@@ Recording failure ' + meth + ' original ' + options.original + '( as ' + options.resource + ') : ' + statusCode) } else { console.log('@@ Recording ' + meth + ' failure for ' + options.original + ': ' + statusCode) } this.requested[Uri.docpart(options.original.uri)] = statusCode this.fireCallbacks('fail', [options.original.uri, errorMessage]) } var err = new Error('Fetcher: ' + errorMessage) // err.ok = false // Is taken as a response, will work too @@ phase out? err.status = statusCode err.statusText = errorMessage err.response = response return Promise.reject(err) } // in the why part of the quad distinguish between HTML and HTTP header // Reverse is set iif the link was rev= as opposed to rel= linkData (originalUri, rel, uri, why, reverse) { if (!uri) return let kb = this.store let predicate // See http://www.w3.org/TR/powder-dr/#httplink for describedby 2008-12-10 let obj = kb.sym(Uri.join(uri, originalUri.uri)) if (rel === 'alternate' || rel === 'seeAlso' || rel === 'meta' || rel === 'describedby') { if (obj.uri === originalUri.uri) { return } predicate = ns.rdfs('seeAlso') } else if (rel === 'type') { predicate = kb.sym('http://www.w3.org/1999/02/22-rdf-syntax-ns#type') } else { // See https://www.iana.org/assignments/link-relations/link-relations.xml // Alas not yet in RDF yet for each predicate // encode space in e.g. rel="shortcut icon" predicate = kb.sym( Uri.join(encodeURIComponent(rel), 'http://www.iana.org/assignments/link-relations/') ) } if (reverse) { kb.add(obj, predicate, originalUri, why) } else { kb.add(originalUri, predicate, obj, why) } } parseLinkHeader (linkHeader, originalUri, reqNode) { if (!linkHeader) { return } // const linkexp = /<[^>]*>\s*(\s*;\s*[^()<>@,;:"/[\]?={} \t]+=(([^()<>@,;:"/[]?={} \t]+)|("[^"]*")))*(,|$)/g // const paramexp = /[^()<>@,;:"/[]?={} \t]+=(([^()<>@,;:"/[]?={} \t]+)|("[^"]*"))/g // From https://www.dcode.fr/regular-expression-simplificator: // const linkexp = /<[^>]*>\s*(\s*;\s*[^()<>@,;:"/[\]?={} t]+=["]))*[,$]/g // const paramexp = /[^\\<>@,;:"\/\[\]?={} \t]+=["])/g // Original: const linkexp = /<[^>]*>\s*(\s*;\s*[^()<>@,;:"/[\]?={} \t]+=(([^\(\)<>@,;:"\/\[\]\?={} \t]+)|("[^"]*")))*(,|$)/g const paramexp = /[^\(\)<>@,;:"\/\[\]\?={} \t]+=(([^\(\)<>@,;:"\/\[\]\?={} \t]+)|("[^"]*"))/g const matches = linkHeader.match(linkexp) for (let i = 0; i < matches.length; i++) { let split = matches[i].split('>') let href = split[0].substring(1) let ps = split[1] let s = ps.match(paramexp) for (let j = 0; j < s.length; j++) { let p = s[j] let paramsplit = p.split('=') // var name = paramsplit[0] let rel = paramsplit[1].replace(/["']/g, '') // '" this.linkData(originalUri, rel, href, reqNode) } } } doneFetch (options, response) { this.addStatus(options.req, 'Done.') this.requested[options.original.uri] = 'done' this.fireCallbacks('done', [options.original.uri]) response.req = options.req // Set the request meta blank node return response } /** * Note two nodes are now smushed * If only one was flagged as looked up, then the new node is looked up again, * which will make sure all the URIs are dereferenced */ nowKnownAs (was, now) { if (this.lookedUp[was.uri]) { // Transfer userCallback if (!this.lookedUp[now.uri]) { this.lookUpThing(now, was) } } else if (this.lookedUp[now.uri]) { if (!this.lookedUp[was.uri]) { this.lookUpThing(was, now) } } } /** * Writes back to the web what we have in the store for this uri * * @param uri {Node|string} * @param [options={}] * * @returns {Promise} */ putBack (uri, options = {}) { uri = uri.uri || uri // Accept object or string let doc = NamedNode.find(uri).doc() // strip off # options.contentType = options.contentType || 'text/turtle' options.data = serialize(doc, this.store, doc.uri, options.contentType) return this.webOperation('PUT', uri, options) } webCopy (here, there, contentType) { return this.webOperation('GET', here) .then((result) => { return this.webOperation( 'PUT', // change to binary from text there, { data: result.responseText, contentType }) }) } /** * @param uri {string} * @param [options] {Object} * * @returns {Promise<Response>} */ delete (uri, options) { return this.webOperation('DELETE', uri, options) .then(response => { this.requested[uri] = 404 this.nonexistent[uri] = true this.unload(this.store.sym(uri)) return response }) } /** * @param parentURI {string} URI of parent container * @param [folderName] {string} Optional folder name (slug) * @param [data] {string} Optional folder metadata * * @returns {Promise<Response>} */ createContainer (parentURI, folderName, data) { let headers = { // Force the right mime type for containers 'content-type': 'text/turtle', 'link': ns.ldp('BasicContainer') + '; rel="type"' } if (folderName) { headers['slug'] = folderName } let options = { headers } if (data) { options.body = data } return this.webOperation('POST', parentURI, options) } /** * A generic web opeation, at the fetch() level. * does not invole the quadstore. * * Returns promise of Response * If data is returned, copies it to response.responseText before returning * * @param method * @param uri or NamedNode * @param options * * @returns {Promise<Response>} */ webOperation (method, uri, options = {}) { uri = uri.uri || uri // Allow a NamedNode to be passed as it is very common options.method = method options.body = options.data || options.body options.force = true const fetcher = this if (options.body && !options.contentType) { throw new Error('Web operation sending data must have a defined contentType.') } if (options.contentType) { options.headers = options.headers || {} options.headers['content-type'] = options.contentType } if (Fetcher.withCredentials(uri, options)) { options.credentials = 'include' // Otherwise coookies are not sent } return new Promise(function (resolve, reject) { fetcher._fetch(uri, options).then(response => { if (response.ok) { if (response.body) { response.text().then(data => { response.responseText = data resolve(response) }) } else { resolve(response) } } else { let msg = 'Web error: ' + response.status if (response.statusText) msg += ' (' + response.statusText + ')' msg += ' on ' + method + ' of <' + uri + '>' if (response.responseText) msg += ': ' + response.responseText let e2 = new Error(msg) e2.response = response reject(e2) } }, err => { let msg = 'Fetch error for ' + method + ' of <' + uri + '>:' + err reject(new Error(msg)) }) }) } /** * Looks up something. * Looks up all the URIs a things has. * * @param term {NamedNode} canonical term for the thing whose URI is * to be dereferenced * @param rterm {NamedNode} the resource which referred to this * (for tracking bad links) * * @returns {Promise} */ lookUpThing (term, rterm) { let uris = this.store.uris(term) // Get all URIs uris = uris.map(u => Uri.docpart(u)) // Drop hash fragments uris.forEach(u => { this.lookedUp[u] = true }) return this.load(uris, { referringTerm: rterm }) } /** * Looks up response header. * * @param doc * @param header * * @returns {Array|undefined} a list of header values found in a stored HTTP * response, or [] if response was found but no header found, * or undefined if no response is available. * Looks for { [] link:requestedURI ?uri; link:response [ httph:header-name ?value ] } */ getHeader (doc, header) { const kb = this.store const requests = kb.each(undefined, ns.link('requestedURI'), doc.uri) for (let r = 0; r < requests.length; r++) { let request = requests[r] if (request !== undefined) { let response = kb.any(request, ns.link('response')) if (response !== undefined) { console.log('@@@ looking for ' + ns.httph(header.toLowerCase())) let results = kb.each(response, ns.httph(header.toLowerCase())) if (results.length) { return results.map(v => { return v.value }) } return [] } } } return undefined } /** * * @param docuri * @param options */ saveRequestMetadata (docuri, options) { let req = options.req let kb = this.store let rterm = options.referringTerm this.addStatus(options.req, 'Accept: ' + options.headers['accept']) if (rterm && rterm.uri) { kb.add(docuri, ns.link('requestedBy'), rterm.uri, this.appNode) } if (options.original && options.original.uri !== docuri) { kb.add(req, ns.link('orginalURI'), kb.literal(options.original.uri), this.appNode) } const now = new Date() const timeNow = '[' + now.getHours() + ':' + now.getMinutes() + ':' + now.getSeconds() + '] ' kb.add(req, ns.rdfs('label'), kb.literal(timeNow + ' Request for ' + docuri), this.appNode) kb.add(req, ns.link('requestedURI'), kb.literal(docuri), this.appNode) kb.add(req, ns.link('status'), kb.collection(), this.appNode) } saveResponseMetadata (response, options) { const kb = this.store let responseNode = kb.bnode() kb.add(options.req, ns.link('response'), responseNode, responseNode) kb.add(responseNode, ns.http('status'), kb.literal(response.status), responseNode) kb.add(responseNode, ns.http('statusText'), kb.literal(response.statusText), responseNode) if (!options.resource.uri.startsWith('http')) { return responseNode } // Save the response headers response.headers.forEach((value, header) => { kb.add(responseNode, ns.httph(header), kb.literal(value), responseNode) if (header === 'content-type') { kb.add(options.resource, ns.rdf('type'), Util.mediaTypeClass(value), responseNode) } }) return responseNode } objectRefresh (term) { let uris = this.store.uris(term) // Get all URIs if (typeof uris !== 'undefined') { for (let i = 0; i < uris.length; i++) { this.refresh(this.store.sym(Uri.docpart(uris[i]))) // what about rterm? } } } /* refresh Reload data from a given document ** ** @param {NamedNode} term - An RDF Named Node for the eodcument in question ** @param {function } userCallback - A function userCallback(ok, message, response) */ refresh (term, userCallback) { // sources_refresh this.fireCallbacks('refresh', arguments) this.nowOrWhenFetched(term, { force: true, clearPreviousData: true }, userCallback) } /* refreshIfExpired Conditional refresh if Expired ** ** @param {NamedNode} term - An RDF Named Node for the eodcument in question ** @param {function } userCallback - A function userCallback(ok, message, response) */ refreshIfExpired (term, userCallback) { let exp = this.getHeader(term, 'Expires') if (!exp || (new Date(exp).getTime()) <= (new Date().getTime())) { this.refresh(term, userCallback) } else { userCallback(true, 'Not expired', {}) } } retract (term) { // sources_retract this.store.removeMany(undefined, undefined, undefined, term) if (term.uri) { delete this.requested[Uri.docpart(term.uri)] } this.fireCallbacks('retract', arguments) } getState (docuri) { if (typeof this.requested[docuri] === 'undefined') { return 'unrequested' } else if (this.requested[docuri] === true) { return 'requested' } else if (this.requested[docuri] === 'done') { return 'fetched' } else if (this.requested[docuri] === 'redirected') { return this.getState(this.redirectedTo[docuri]) } else { // An non-200 HTTP error status return 'failed' } } isPending (docuri) { // sources_pending // doing anyStatementMatching is wasting time // if it's not pending: false -> flailed // 'done' -> done 'redirected' -> redirected return this.requested[docuri] === true } unload (term) { this.store.removeDocument(term) delete this.requested[term.uri] // So it can be loaded again } addHandler (handler) { this.handlers.push(handler) handler.register(this) } retryNoCredentials (docuri, options) { console.log('web: Retrying with no credentials for ' + options.resource) options.retriedWithNoCredentials = true // protect against being called twice delete this.requested[docuri] // forget the original request happened let newOptions = Object.assign({}, options, { withCredentials: false }) this.addStatus(options.req, 'Abort: Will retry with credentials SUPPRESSED to see if that helps') return this.load(docuri, newOptions) } /** * Tests whether a request is being made to a cross-site URI (for purposes * of retrying with a proxy) * * @param uri {string} * * @returns {boolean} */ isCrossSite (uri) { // Mashup situation, not node etc if (typeof document === 'undefined' || !document.location) { return false } const hostpart = Uri.hostpart const here = '' + document.location return hostpart(here) && hostpart(uri) && hostpart(here) !== hostpart(uri) } /** * Called when there's a network error in fetch(), or a response * with status of 0. * * @param response {Response|Error} * @param docuri {string} * @param options {Object} * * @returns {Promise} */ handleError (response, docuri, options) { if (this.isCrossSite(docuri)) { // Make sure we haven't retried already if (options.withCredentials && !options.retriedWithNoCredentials) { return this.retryNoCredentials(docuri, options) } // Now attempt retry via proxy let proxyUri = Fetcher.crossSiteProxy(docuri) if (proxyUri && !options.proxyUsed) { console.log('web: Direct failed so trying proxy ' + proxyUri) return this.redirectToProxy(proxyUri, options) } } var message if (response.message) { message = 'Fetch error: ' + response.message } else { message = response.statusText if (response.responseText) { message += ` ${response.responseText}` } } // This is either not a CORS error, or retries have been made return this.failFetch(options, message, response.status || 998, response) } // deduce some things from the HTTP transaction addType (rdfType, req, kb, locURI) { // add type to all redirected resources too let prev = req if (locURI) { var reqURI = kb.any(prev, ns.link('requestedURI')) if (reqURI && reqURI !== locURI) { kb.add(kb.sym(locURI), ns.rdf('type'), rdfType, this.appNode) } } for (;;) { const doc = kb.any(prev, ns.link('requestedURI')) if (doc && doc.value) { kb.add(kb.sym(doc.value), ns.rdf('type'), rdfType, this.appNode) } // convert Literal prev = kb.any(undefined, kb.sym('http://www.w3.org/2007/ont/link#redirectedRequest'), prev) if (!prev) { break } var response = kb.any(prev, kb.sym('http://www.w3.org/2007/ont/link#response')) if (!response) { break } var redirection = kb.any(response, kb.sym('http://www.w3.org/2007/ont/http#status')) if (!redirection) { break } if (redirection !== '301' && redirection !== '302') { break } } } /** * Handle fetch() response * * @param response {Response} fetch() response object * @param docuri {string} * @param options {Object} */ handleResponse (response, docuri, options) { const kb = this.store const headers = response.headers const reqNode = options.req const responseNode = this.saveResponseMetadata(response, options) const contentType = this.normalizedContentType(options, headers) || '' let contentLocation = headers.get('content-location') // this.fireCallbacks('recv', xhr.args) // this.fireCallbacks('headers', [{uri: docuri, headers: xhr.headers}]) // Check for masked errors (CORS, etc) if (response.status === 0) { console.log('Masked error - status 0 for ' + docuri) return this.handleError(response, docuri, options) } if (response.status >= 400) { if (response.status === 404) { this.nonexistent[options.original.uri] = true this.nonexistent[docuri] = true } return this.saveErrorResponse(response, responseNode) .then(() => { let errorMessage = options.resource + ' ' + response.statusText return this.failFetch(options, errorMessage, response.status, response) }) } var diffLocation = null var absContentLocation = null if (contentLocation) { absContentLocation = Uri.join(contentLocation, docuri) if (absContentLocation !== docuri) { diffLocation = absContentLocation } } if (response.status === 200) { this.addType(ns.link('Document'), reqNode, kb, docuri) if (diffLocation) { this.addType(ns.link('Document'), reqNode, kb, diffLocation) } // Before we parse new data clear old but only on 200 if (options.clearPreviousData) { kb.removeDocument(options.resource) } let isImage = contentType.includes('image/') || contentType.includes('application/pdf') if (contentType && isImage) { this.addType(kb.sym('http://purl.org/dc/terms/Image'), reqNode, kb, docuri) if (diffLocation) { this.addType(kb.sym('http://purl.org/dc/terms/Image'), reqNode, kb, diffLocation) } } } // If we have already got the thing at this location, abort if (contentLocation) { if (!options.force && diffLocation && this.requested[absContentLocation] === 'done') { // we have already fetched this // should we smush too? // log.info("HTTP