@fastly/as-url/assembly/url-parse.ts

// Copyright 2021 Fastly, Inc.

import { SPECIAL_SCHEMES, throwInvalidUrlError } from "./util";

import { URLProperties } from "./url-properties";

export class URLParser {
  static isAbsoluteUrl(url: string): boolean {
    if (url.startsWith("//")) {
      // Protocol Relative URL
      return false;
    }
    if (url.indexOf(":") > 0) {
      return true;
    }

    return false;
  }

  static applySchemeOrPathRelativeUrl(
    relativeUrl: string,
    urlProps: URLProperties
  ): void {
    // We already have our absolute URL when this is called
    // So we just need to work off the existing properties

    // Figure out our type of relative URL
    // https://url.spec.whatwg.org/#relative-url-string

    if (relativeUrl.startsWith("//")) {
      // Protocol (scheme) relative URLs
      let urlAfterAuth = URLParser.parseAuth(
        relativeUrl.substring(2),
        urlProps
      );
      let urlAfterHost = URLParser.parseHost(urlAfterAuth, urlProps);
      let urlAfterPath = URLParser.parsePath(urlAfterHost, urlProps);
      let urlAfterSearch = URLParser.parseSearch(urlAfterPath, urlProps);
      URLParser.parseHash(urlAfterSearch, urlProps);
      return;
    }

    // Check if we need to preserve the original pathname
    if (relativeUrl.startsWith(".")) {
      // Do Nothing
    } else {
      urlProps.pathname = "";
      urlProps.search = "";
      urlProps.hash = "";
    }

    // Must be a path relative URL
    URLParser.applyPathRelativeUrl(relativeUrl, urlProps);
  }

  static applyPathRelativeUrl(
    relativeUrl: string,
    urlProps: URLProperties
  ): void {
    // We already have our absolute URL when this is called
    // So we just need to work off the existing properties

    // Handle path navigation in path relative URLs (e.g, ../ or ./) or path absolute urls (e.g, /path/absolute)
    let appliedRelativeUrl = urlProps.pathname;
    if (relativeUrl.startsWith("/")) {
      appliedRelativeUrl = relativeUrl;
    } else {
      if (!appliedRelativeUrl.endsWith("/") && !relativeUrl.startsWith("/")) {
        appliedRelativeUrl += "/" + relativeUrl;
      } else {
        appliedRelativeUrl += relativeUrl;
      }
    }

    // Ensure our navigation identifiers have the correct slashes
    if (appliedRelativeUrl.endsWith(".")) {
      appliedRelativeUrl += "/";
    }

    // Remove any filler navigation (e.g ./)
    while (appliedRelativeUrl.includes("/./")) {
      appliedRelativeUrl = appliedRelativeUrl.replace("/./", "/");
    }

    // Do any parent navigation
    while (appliedRelativeUrl.includes("/../")) {
      let parentDirectoryIndex = appliedRelativeUrl.indexOf("../");

      // > 1 because the leading slash will be there
      if (parentDirectoryIndex > 1) {
        // Remove this directory, and the one before
        let parentIndex = appliedRelativeUrl.lastIndexOf(
          "/",
          parentDirectoryIndex - 2
        );
        let parentReplaceTerm = appliedRelativeUrl.slice(
          parentIndex,
          parentDirectoryIndex + 3
        );

        appliedRelativeUrl = appliedRelativeUrl.replace(parentReplaceTerm, "/");
      } else {
        throw new Error(
          "Relative url " +
            relativeUrl +
            " cannot be applied to the url " +
            urlProps.toString()
        );
      }
    }

    relativeUrl = appliedRelativeUrl;
    if (!relativeUrl.startsWith("/")) {
      relativeUrl = "/" + relativeUrl;
    }

    // We hit a normal path relative URL (e.g hello/goodbye/)
    // Just continue on our current path
    let urlAfterHost = relativeUrl + urlProps.search + urlProps.hash;
    let urlAfterPath = URLParser.parsePath(urlAfterHost, urlProps);
    let urlAfterSearch = URLParser.parseSearch(urlAfterPath, urlProps);
    URLParser.parseHash(urlAfterSearch, urlProps);

    // Also, if our original relative URL had a trailing slash, we need to re-add that trailing slash
    if (relativeUrl.endsWith("/") && !urlProps.pathname.endsWith("/")) {
      urlProps.pathname += "/";
    }

    return;
  }

  static parseAbsoluteUrl(absoluteUrl: string, urlProps: URLProperties): void {
    // For file URLs, we should replace | with :
    if (absoluteUrl.startsWith("file:")) {
      absoluteUrl = absoluteUrl.replaceAll("|", ":");
    }

    // Chain our independent parsing functions
    let urlAfterProtocol = URLParser.parseProtocol(absoluteUrl, urlProps);
    let urlAfterAuth = URLParser.parseAuth(urlAfterProtocol, urlProps);
    let urlAfterHost = URLParser.parseHost(urlAfterAuth, urlProps);

    // Now we are at the path, let's apply the relative URL on top of our empty path
    if (urlAfterHost.length > 0) {
      urlProps.pathname = "";
      urlProps.search = "";
      urlProps.hash = "";

      // Lets remove all parent navigation as they are ignored on absolute URLs
      while (urlAfterHost.startsWith("/../")) {
        urlAfterHost = urlAfterHost.replace("/../", "/");
      }

      URLParser.applyPathRelativeUrl(urlAfterHost, urlProps);
    }
  }

  // Takes in an absolute URL,
  // Apply the URL protocol from the absolute url
  // and returns a partial url with everything after the protocol (auth, host, pathname, search, hash).
  static parseProtocol(absoluteUrl: string, urlProps: URLProperties): string {
    // Get the protocol and remaining URL
    let protocolIndex = absoluteUrl.indexOf(":");

    // Find where the slashes end after the :
    if (protocolIndex > -1) {
      urlProps.protocol = absoluteUrl.substring(0, protocolIndex + 1);

      // Files need to have :// , and the path starts at the third slash
      // Ignore anything in between
      if (urlProps.protocol == "file:") {
        let absoluteUrlNoProtocol = absoluteUrl.replace(
          urlProps.protocol + "//",
          ""
        );

        if (absoluteUrlNoProtocol.indexOf("/") > -1) {
          return absoluteUrlNoProtocol.substring(
            absoluteUrlNoProtocol.indexOf("/")
          );
        } else {
          return "";
        }
      }

      let protocolEndIndex = protocolIndex + 1;
      while (
        absoluteUrl.charAt(protocolEndIndex) == "/" &&
        protocolEndIndex < absoluteUrl.length - 1
      ) {
        protocolEndIndex++;
      }

      return absoluteUrl.substring(protocolEndIndex);
    }

    // Did not have a protocol
    return absoluteUrl;
  }

  // Takes in a partial URL without the protocol
  // Applies auth from the partial URL (if there is one)
  // Returns a partial URL with everything after auth (host, pathname, search, hash).
  static parseAuth(urlAfterProtocol: string, urlProps: URLProperties): string {
    // Next, try to get a username and password
    let authIndex = urlAfterProtocol.indexOf("@");
    // This will require finding out the remaining pieces of the url
    // We can check for the path, because the host wont neccesariy have a
    // '.' for the domain (localhost) or a ':' for the port
    let pathIndex = urlAfterProtocol.indexOf("/");
    // @ must proceed the path (/) as stated in the W3 URI Spec
    if (authIndex > 0 && (pathIndex == -1 || authIndex < pathIndex)) {
      let auth = urlAfterProtocol.substring(0, authIndex);

      if (auth.includes(":")) {
        let authSplit = auth.split(":");
        urlProps.username = authSplit[0];
        urlProps.password = authSplit[1];
      } else {
        urlProps.username = auth;
      }

      // Return the remaining url
      return urlAfterProtocol.substring(auth.length + 1);
    }

    // If there was no auth, just return the url
    return urlAfterProtocol;
  }

  // Takes in a partial URL without the protocol or auth
  // Applies the host from the partial url
  // Returns a partial URL with everything after the host (pathname, search, hash).
  static parseHost(urlAfterAuth: string, urlProps: URLProperties): string {
    // Create our return value
    let urlAfterHost = "";

    // Next let's get the hostname and port
    // This will require finding out the remaining pieces of the url
    let hostnameAndPort = "";
    let pathIndex = urlAfterAuth.indexOf("/");
    let searchIndex = urlAfterAuth.indexOf("?");
    let hashIndex = urlAfterAuth.indexOf("#");
    if (pathIndex > -1) {
      hostnameAndPort = urlAfterAuth.substring(0, pathIndex);
      urlAfterHost = urlAfterAuth.substring(pathIndex);
    } else if (searchIndex > -1) {
      hostnameAndPort = urlAfterAuth.substring(0, searchIndex);
      urlAfterHost = urlAfterAuth.substring(searchIndex);
    } else if (hashIndex > -1) {
      hostnameAndPort = urlAfterAuth.substring(0, hashIndex);
      urlAfterHost = urlAfterAuth.substring(hashIndex);
    } else {
      hostnameAndPort = urlAfterAuth;
      urlAfterHost = "";
    }

    let hostname = "";
    let port = "";

    if (hostnameAndPort.includes("[")) {
      // This could be an ipv6 address
      // https://url.spec.whatwg.org/#host-writing

      if (!hostnameAndPort.startsWith("[") || !hostnameAndPort.endsWith("]")) {
        throwInvalidUrlError();
      }

      let splitAddress = hostnameAndPort.split(":");
      if (splitAddress.length != 8) {
        throwInvalidUrlError();
      }

      hostname = hostnameAndPort;
    } else if (hostnameAndPort.includes(":")) {
      let hostnameAndPortSplit = hostnameAndPort.split(":");
      hostname = hostnameAndPortSplit[0];

      // The port must be a number, so try to parse it
      let portOrNaN = F32.parseInt(hostnameAndPortSplit[1], 10);
      if (isNaN(portOrNaN) || portOrNaN <= 0 || portOrNaN >= 65536) {
        throwInvalidUrlError();
      }

      port = I32.parseInt(hostnameAndPortSplit[1], 10).toString();
    } else {
      hostname = hostnameAndPort;
    }

    // Ensure that port is null if the port is a default port per the protocol
    // https://url.spec.whatwg.org/#default-port
    if (
      port.length > 0 &&
      ((urlProps.protocol == "ftp:" && port == "22") ||
        (urlProps.protocol == "http:" && port == "80") ||
        (urlProps.protocol == "https:" && port == "443") ||
        (urlProps.protocol == "ws:" && port == "80") ||
        (urlProps.protocol == "wss:" && port == "443"))
    ) {
      port = "";
    }

    urlProps.hostname = hostname;
    urlProps.port = port;

    // return our resulting URL
    return urlAfterHost;
  }

  // Takes in a parital URL without the protocol, auth, or host
  // Applies the path from the partial url
  // Returns a partial URL with everything after the path (search, hash).
  static parsePath(urlAfterHost: string, urlProps: URLProperties): string {
    if (urlAfterHost.length == 0) {
      return "";
    }

    let pathIndex = urlAfterHost.indexOf("/");
    let searchIndex = urlAfterHost.indexOf("?");
    let hashIndex = urlAfterHost.indexOf("#");

    if (pathIndex > -1) {
      // Get the pathname
      if (searchIndex > -1) {
        urlProps.pathname = urlAfterHost.substring(0, searchIndex);
      } else if (hashIndex > -1) {
        urlProps.pathname = urlAfterHost.substring(0, hashIndex);
      } else {
        urlProps.pathname = urlAfterHost;
      }

      // Remove any trailing slash, if the character before is not a slash
      if (
        urlProps.pathname.endsWith("/") &&
        !urlProps.pathname.endsWith("//")
      ) {
        urlProps.pathname = urlProps.pathname.slice(
          0,
          urlProps.pathname.length - 1
        );
      }

      // Return the reamaining string without the path
      if (searchIndex > -1) {
        return urlAfterHost.substring(searchIndex);
      } else if (hashIndex > -1) {
        return urlAfterHost.substring(hashIndex);
      } else {
        return "";
      }
    }

    // Just return the string if there was no path
    return urlAfterHost;
  }

  // Takes in a partial URL without the protocol, auth, host, or path
  // Applies the search from the partial url
  // Returns a partial URL with everything after the search (hash).
  static parseSearch(urlAfterPath: string, urlProps: URLProperties): string {
    if (urlAfterPath.length == 0) {
      return "";
    }

    let searchIndex = urlAfterPath.indexOf("?");
    let hashIndex = urlAfterPath.indexOf("#");
    if (searchIndex > -1) {
      if (hashIndex > -1) {
        urlProps.search = urlAfterPath.substring(0, hashIndex);
        return urlAfterPath.substring(hashIndex);
      } else {
        urlProps.search = urlAfterPath;
        return "";
      }
    }

    // Just return the string if there was no search
    return urlAfterPath;
  }

  // Takes in a partial URL without the protocol, auth, host, path, or search (Only a hash)
  // Applies the hash from the partial url
  static parseHash(urlAfterSearch: string, urlProps: URLProperties): void {
    let hashIndex = urlAfterSearch.indexOf("#");
    if (urlAfterSearch.length > 0 && hashIndex > -1) {
      urlProps.hash = urlAfterSearch.substring(hashIndex);
    }
  }

  static validateUrl(urlProps: URLProperties): void {
    // Hostname checks

    if (urlProps.hostname.includes(".")) {
      // Check if the hostname is a domain or IPv4 address

      if (!SPECIAL_SCHEMES.includes(urlProps.protocol)) {
        // This is an invlid URL according to the spec:
        // https://url.spec.whatwg.org/#url-representation
        // However, this is supported by node and chrome:
        // https://nodejs.org/api/url.html#url_special_schemes
        // Do Nothing, instead of throwing the error below:
        // throw new Error("Failed to construct 'URL': Invalid URL");
      }
    } else if (urlProps.hostname == "") {
      // Check for empty host

      if (
        SPECIAL_SCHEMES.includes(urlProps.protocol) &&
        urlProps.protocol != "file:"
      ) {
        throwInvalidUrlError();
      }
    } else {
      // Must be an opaque host (e.g localhost), or ipv6
      if (urlProps.hostname.includes("[")) {
        // This url is ipv6, we are good!
      } else if (
        SPECIAL_SCHEMES.includes(urlProps.protocol) &&
        urlProps.protocol != "http:" &&
        urlProps.protocol != "https:"
      ) {
        // We do not want to allow special schemes for opaque hosts,
        // but for opaque hosts like localhost, http: and https: is valid in v8.
        // Thus, we should allow those, but not other special schemes.
        throwInvalidUrlError();
      }
    }

    // The Url is valid!
  }
}