Basic HTTP support and URL parsing

4 files changed, 312 insertions, 0 deletions

diff --git a/src/browser.nim b/src/browser.nim
new file mode 100644
index 0000000..97713a3
--- /dev/null
+++ b/src/browser.nim
@@ -0,0 +1,33 @@
+import std/strutils, protocols/http, html, uri
+
+let url = "https://example.org:443/index.html"
+# let url = paramStr(1)
+
+proc request(uri: string) =
+  # This is probably the best place to implement scheme-specific stuff
+  let url = parseURL(uri)
+  case url.scheme:
+    of "http", "https":
+      let response = httpRequest(url)
+      renderHTML(response.body)
+    # Exercise: view-source
+    of "view-source":
+      # We must parse the url again without the view-source: prefix
+      let url = uri.split(':', maxsplit=1)[1]
+      let response = httpRequest(parseURL(url))
+      renderSource(response.body)
+    # Exercise: file:// scheme
+    of "file":
+      discard
+    # Exercise: data scheme
+    of "data":
+      discard
+    else:
+      raise newException(Exception, "Not a valid scheme: " & url.scheme)
+
+request(url)
+
+# HTTP/1.1
+# Compression
+# Redirects
+# Caching
diff --git a/src/html.nim b/src/html.nim
new file mode 100644
index 0000000..4ae5ee1
--- /dev/null
+++ b/src/html.nim
@@ -0,0 +1,30 @@
+type Html = object
+  tags: seq[string]
+
+type Tag {.acyclic.} = object
+  name: string
+  text: string
+  nested: seq[Tag]
+
+func parseHTML(html: string): Html = discard
+
+# Todo: revamp parsing: keep track of tags, entities, etc
+proc renderHTML*(html: string) =
+  var
+    in_angle = false
+    in_body = false
+
+  # _Why_ is it i, c and not c, i...
+  for i, c in html:
+    if c == '<':
+      in_angle = true
+      if html[i..i+4] == "<body":
+        in_body = true
+    elif c == '>':
+      in_angle = false
+    elif not in_angle and in_body and c in {char(32)..char(126), '\n'}:
+      stdout.write(c)
+
+proc renderSource*(html: string) =
+  for i, c in html:
+    stdout.write(c)
diff --git a/src/protocols/http.nim b/src/protocols/http.nim
new file mode 100644
index 0000000..df812fd
--- /dev/null
+++ b/src/protocols/http.nim
@@ -0,0 +1,67 @@
+import std/[strutils, net], ../uri
+
+# https://datatracker.ietf.org/doc/html/rfc1945
+
+type Http* = object of RootObj
+  version*: string # always present
+  headers*: seq[tuple[header: string, value: string]]
+
+type HttpRequest* = object of Http
+  `method`*: string  # only present in requests
+  uri*: string       # only present in requests
+
+type HttpResponse* = object of Http
+  status*: int     # status code, only in responses
+  reason*: string  # status elaboration, only in responses
+  body*: string    # html document, usually
+
+# This parses a HTTP response that has been split into headers and a body.
+func parseResponse*(http: seq[string], body: string): HttpResponse =
+  # let http: seq[string] = http.split("\r\n")
+  let split = http[0].split(' ', maxsplit=2)
+  if split.len == 3:
+    result.version = split[0]
+    result.status = split[1].parseInt
+    result.reason = split[2]
+  else:
+    raise newException(RangeDefect, "First line of response is invalid: " & http[0])
+  # Note: the spec specifies that \r\n\r\n marks the end of a request
+  for header in http[1 ..< http.len]:
+    let split = header.split(':', maxsplit=1)
+    if split.len == 2:
+      result.headers.add((split[0].toLower, split[1].strip()))
+  result.body = body
+
+proc httpRequest*(url: Url): HttpResponse =
+  let exampleRequest = "GET " & url.path & " HTTP/1.0\r\nHost: example.com\r\n\r\n"
+
+  let socket = newSocket(AF_INET, SOCK_STREAM, IPPROTO_TCP)
+
+  var port = 80
+  if url.scheme == "https":
+    port = 443
+    let ctx: SSLContext = newContext()
+    ctx.wrapSocket(socket)
+  if url.port != 0:
+    port = url.port
+
+  socket.connect(url.host, Port(port))
+  socket.send(exampleRequest)
+
+  var
+    response: seq[string]
+    buffer: string
+  while true:
+    socket.readLine(buffer)
+    if buffer == "\r\n":
+      break
+    response.add(buffer)
+
+  # assert "transfer-encoding" notin parsed.headers
+  # assert "content-encoding" notin parsed.headers
+
+  var body: string
+  while socket.recv(buffer, 1024) > 0: # why?
+    body &= buffer
+
+  result = parseResponse(response, body)
diff --git a/src/uri.nim b/src/uri.nim
new file mode 100644
index 0000000..e2fc7f3
--- /dev/null
+++ b/src/uri.nim
@@ -0,0 +1,182 @@
+import std/strutils
+
+# https://datatracker.ietf.org/doc/html/rfc3986
+type Url* = object
+  scheme*: string    # :
+  authority*: string # //
+  userinfo*: string  # @
+  host*: string      # .
+  port*: int         # :
+  path*: string      # /
+  query*: string     # ?
+  fragment*: string  # #
+
+const
+  gendelims* = {':', '/', '?', '#', '[', ']', '@'}
+  subdelims* = {'!', '$', '&', '\'', '(', ')', '*', '+', ',', ';', '='}
+  reserved* = gendelims + subdelims
+  unreserved* = {'a'..'z', 'A'..'Z', '0'..'9', '-', '.', '_', '~'}
+  otherchars* = {'"', '%', '<', '>', '\\', '^', '`', '{', '|', '}', } # plus, whitespace and control characters
+
+# This might be spec-compliant.
+func parseURL*(url: string): Url =
+  # Each URI must begin with a scheme name.
+  # Scheme names consist of a sequence of characters beginning with a letter,
+  # and followed by any combination of letters, digits, plus, period, or hyphen.
+  var split = url.split(':', maxsplit=1)
+  if split.len == 2:
+    # Although schemes are case-insensitive, an implementation should produce
+    # lowercase scheme names for consistency.
+    result.scheme = split[0].toLower
+    assert allCharsInSet(result.scheme, {'a'..'z', '0'..'9', '+', '-', '.'})
+
+    # The authority component is preceded by a double slash, and is terminated by
+    # the next slash, question mark, or pound sign, or by the end of the URI.
+    if split[1][0..1] == "//":
+      result.authority = split[1][2..^1].split({'/', '?', '#'}, maxsplit=1)[0]
+      split[1] = split[1][result.authority.len + 2 ..< split[1].len]
+
+    # A fragment identifier component is indicate by the presense of a pound sign
+    # character, and terminated by the end of the URI.
+    split = split[1].split('#', maxsplit=1)
+    if split.len == 2:
+      result.fragment = split[1]
+
+    # The query component is indicated by the first question mark character,
+    # and terminated by a pound sign or the end of the URI.
+    split = split[0].split('?', maxsplit=1)
+    if split.len == 2:
+      result.query = split[1]
+
+    # Todo: path parsing is actually more complex. this is cheating.
+    if split[0] != "":
+      result.path = "/" & split[0].strip(leading=true, trailing=false, {'/'})
+
+    if result.authority != "":
+      # The port subcomponent, if present, is designated by an optional port number
+      # in decimal following the host and deliminated from it by a single colon.
+      split = result.authority.rsplit({':'}, maxsplit=1)
+      if split.len == 2:
+        # If we accidentally parsed an IPv6 address: reconstruct
+        if ']' in split[1]:
+          split[0] = split[0] & ":" & split[1]
+          split[1] = ""
+        else:
+          result.port = split[1].parseInt
+          assert allCharsInSet(split[1], {'0'..'9'})
+
+      # The userinfo subcomponent, if present, is followed by an at-sign
+      # that delimits it from the host.
+      split = split[0].split('@', maxsplit=1)
+      if split.len == 2:
+        result.userinfo = split[0]
+        # Todo: parse hosts
+        result.host = split[1]
+      else:
+        result.host = split[0]
+      # Todo: is this assert right?
+      assert allCharsInSet(result.host, unreserved + subdelims + {':', '[', ']'})
+  else:
+    raise newException(RangeDefect, "URL does not begin with a scheme")
+
+  #[ Todo: advanced resolution and relative references
+  # A relative reference takes advantage of the hierarchical syntax to express
+  # a reference relative to the name space of another hierarchical URI.
+  if url.len == 0:
+    raise newException(RangeDefect, "URL is an empty string")
+  # Relative path reference
+  elif url[0] == '.':
+    return
+  # Network path reference
+  elif url.len > 1 and url[0..1] == "//":
+    return
+  # Absolute path reference
+  elif url[0] == '/':
+    discard url.split('/', maxsplit=1)
+  ]#
+
+#[
+# This is compact. But it's probably not spec-compliant.
+func parseURL*(url: string): Url =
+  # https://google.com/path?query#fragment
+  # https, //google.com/path?query#fragment
+  var split: seq[string] = url.split(':', maxsplit=1)
+  if split.len == 2:
+    result.scheme = split[0]
+
+    # //google.com/path?query, fragment
+    split = split[1].rsplit('#', maxsplit=1)
+    if split.len == 2:
+      result.fragment = split[1]
+
+    # //google.com/path, query
+    # //google.com/, query
+    split = split[0].rsplit('?', maxsplit=1)
+    if split.len == 2:
+      result.query = split[1]
+
+    # //google.com/path
+    # //google.com
+    # somethingelse
+    if split[0][0..1] == "//":
+      split = split[0].strip(leading=true, trailing=false, {'/'}).split('/', maxsplit=1)
+      result.authority = split[0]
+      if split.len == 2:
+        result.path = split[1]
+    else:
+      # Note: This is a weird state.
+      result.path = split[0].strip(leading=true, trailing=true, {'/'})
+      debugEcho split[0]
+
+    # userinfo@host:port
+    # userinfo, host:port
+    var remainder = result.authority
+    split = result.authority.split('@', maxsplit=1)
+    if split.len == 2:
+      result.userinfo = split[0]
+      remainder = split[1]
+
+    # host, port
+    split = remainder.rsplit(':', maxsplit=1)
+    remainder = split[0]
+    if split.len == 2:
+      result.port = split[1].parseInt
+
+    result.host = split[0]
+
+  else:
+    raise newException(OSError, "URL does not contain a colon")
+]#
+
+# assert parseURL("google.com") == Url(authority: "google.com", host: "google.com")
+# assert parseURL("http://") == Url(scheme: "http")
+assert parseURL("https://google.com") == Url(scheme: "https", authority: "google.com", host: "google.com")
+assert parseURL("https://google.com/path//path") == Url(scheme: "https", authority: "google.com", host: "google.com", path: "/path//path")
+assert parseURL("https://google.com/path?query") == Url(scheme: "https", authority: "google.com", host: "google.com", path: "/path", query: "query")
+assert parseURL("https://google.com/path/?query") == Url(scheme: "https", authority: "google.com", host: "google.com", path: "/path/", query: "query")
+assert parseURL("https://google.com?query") == Url(scheme: "https", authority: "google.com", host: "google.com", query: "query")
+assert parseURL("https://google.com/path#fragment") == Url(scheme: "https", authority: "google.com", host: "google.com", path: "/path", fragment: "fragment")
+assert parseURL("https://google.com/path#fragment?query") == Url(scheme: "https", authority: "google.com", host: "google.com", path: "/path", fragment: "fragment?query")
+assert parseURL("https://192.168.0.1/path#fragment?query") == Url(scheme: "https", authority: "192.168.0.1", host: "192.168.0.1", path: "/path", fragment: "fragment?query")
+assert parseURL("https://12:12:12:12/path#fragment?query") == Url(scheme: "https", authority: "12:12:12:12", host: "12:12:12", port: 12, path: "/path", fragment: "fragment?query")
+assert parseURL("https://[12:213:213:fr:12]:1200/path#fragment?query") == Url(scheme: "https", authority: "[12:213:213:fr:12]:1200", host: "[12:213:213:fr:12]", port: 1200, path: "/path", fragment: "fragment?query")
+assert parseURL("https://userinfo@host.com:8080/path#fragment?query") == Url(scheme: "https", authority: "userinfo@host.com:8080", userinfo: "userinfo", host: "host.com", port: 8080, path: "/path", fragment: "fragment?query")
+
+assert parseURL("ftp://ftp.is.co.za/rfc/rfc1808.txt") == Url(scheme: "ftp", authority: "ftp.is.co.za", host: "ftp.is.co.za", path: "/rfc/rfc1808.txt")
+assert parseURL("http://www.ietf.org/rfc/rfc2396.txt") == Url(scheme: "http", authority: "www.ietf.org", host: "www.ietf.org", path: "/rfc/rfc2396.txt")
+assert parseURL("ldap://[2001:db8::7]/c=GB?objectClass?one") == Url(scheme: "ldap", authority: "[2001:db8::7]", host: "[2001:db8::7]", path: "/c=GB", query: "objectClass?one")
+
+# Todo: Ugh, are these right?
+assert parseURL("mailto:John.Doe@example.com") == Url(scheme: "mailto", path: "/John.Doe@example.com") # ugh is this right
+assert parseURL("news:comp.infosystems.www.servers.unix") == Url(scheme: "news", path: "/comp.infosystems.www.servers.unix")
+assert parseURL("tel:+1-816-555-1212") == Url(scheme: "tel", path: "/+1-816-555-1212")
+assert parseURL("telnet://192.0.2.16:80/") == Url(scheme: "telnet", authority: "192.0.2.16:80", host: "192.0.2.16", port: 80, path: "/")
+assert parseURL("urn:oasis:names:specification:docbook:dtd:xml:4.1.2") == Url(scheme: "urn", path: "/oasis:names:specification:docbook:dtd:xml:4.1.2")
+
+func encodeURL(url: string): string =
+  for c in url:
+    if c in unreserved:
+      result &= c
+    # Note: `+`?
+    else:
+      result &= "%" & toHex(ord(c), 2)