HTTPRequestclasshttpreq.h[29], webui.t[101]

HTTP Request object. This object represents an HTTP protocol request from a client to one of our servers. HTTPRequest objects are created by the HTTPServer object as requests arrive, and are passed to the byte code program as network events, via the getNetEvent() function. The program uses the object to get information on the request and to send back the reply.

Modified in webui.t[101]:
HTTPRequest extensions

intrinsic class HTTPRequest :   Object

Superclass Tree   (in declaration order)


Subclass Tree  


Global Objects  


Summary of Properties  


Summary of Methods  

endChunkedReply  getBody  getClientAddress  getCookie  getCookies  getFormFields  getHeaders  getQuery  getQueryParam  getServer  getVerb  parseQuery  sendReply  sendReplyAsync  sendReplyCatch  sendReplyChunk  setCookie  startChunkedReply 

Inherited from Object :
getPropList  getPropParams  getSuperclassList  isClass  isTransient  ofKind  propDefined  propInherited  propType  valToSymbol 




endChunkedReply (headers?)httpreq.h[340]

Finish a chunked reply. This completes a chunked reply started with startChunkedReply(). After calling this routine, the request is completed, and no further reply can be sent.

'headers' is an optional list of additional headers to send with the end of the reply. The HTTP chunked reply mechanism allows a server to send headers at the beginning of the reply (which corresponds to the startChunkedReply() call), at the end of the reply (in this routine), or both. This accommodates situations where the server might not be able to determine the final value for a header until after generating full reply. If this argument is included, it works just like the 'headers' argument to sendReply().

getBody ( )httpreq.h[184]
Get the body of the request, if any. Some types of HTTP requests, such as POST and PUT, contain a message body. This returns the raw, unparsed message body. Returns a File object, open with read-only access. If there's no message body at all, this returns nil.

If the body is a text type, the file will be open in text mode, with the character mapping set according to the content type passed from the client; otherwise it's open in raw binary mode.

The message body retrieved here is the raw request payload, without any processing. There are two standard structured body types that are frequently used with POSTs, both of which require further parsing. First is MIME type application/x-www-form-urlencoded, which is used to represent a basic HTML form, and encodes the data entry fields in a form; this type can be parsed via the getFormFields() method. Second is multipart/form-data, which is used to POST forms that include uploaded files; this can be parsed with getFormFields() to retrieve the data fields, and getUploads() to get the uploaded file data.

getClientAddress ( )httpreq.h[193]
Get the network address of the client. This returns a list: ['ip-address', port], where 'ip-address' is a string with the IP address of the client, in decimal notation ('', for example), and 'port' is an integer giving the network port number on the client side.

getCookie (name)httpreq.h[123]
Look up a cookie. This looks for the given cookie name in the cookies sent by the client, and returns a string containing the cookie's text if found. If the cookie isn't found, returns nil.

getCookies ( )httpreq.h[132]
Get the cookies sent with the request by the client. This returns a LookupTable of the cookies, with each key set to a cookie name and the corresponding value set to the cookie's text. Cookies are assumed to contain only plan ASCII characters; any 8-bit characters in a cookie's name or value will be replaced by '?' characters.

getFormFields ( )httpreq.h[161]
Get the form data-entry field values. This returns a LookupTable containing the field values sent with the request. Each key in the table is a field name (given by the NAME property of the HTML <INPUT> tag for the field), and each corresponding value is a string giving the value of the field as entered by the user.

If the form includes uploaded files (via <INPUT TYPE=FILE>), the value of each file field will be a FileUpload object instead of a string. The 'file' property of this object contains a File object, opened in read-only mode, that can be used to retrieve the contents of the uploaded file. Other properties of object provide the client-side name of the file and the MIME type specified by the browser.

This method is a convenience function that parses the message body information. You can obtain the raw information on the posted data via the getHeader() method.

This method recognizes form data as a message body with content-type application/x-www-form-urlencoded or multipart/form-data. If the request doesn't have a message body at all, or has a message body with a different content-type, this method assumes that the request has no posted form data and returns nil. We return nil rather than an empty lookup table so that the caller can tell that this doesn't appear to be a form submission request at all.

getHeaders ( )httpreq.h[116]
Get the headers sent with the request by the client. This returns a LookupTable object: the keys are the header names, and the values are the corresponding header values. For example, for a POST, there might be a 'Content-type' key with the corresponding value 'application/x-www-form-urlencoded'.

The element table[1] contains the HTTP request line - the first line of the request, which isn't technically a header. This line contains the verb, the query string, and (optionally) the HTTP version string (in that order, with spaces separating the elements). The method includes this so that you can inspect the unparsed request line, if desired.

getQuery ( )httpreq.h[60]
Get the query string. This is the portion of the URL after the server address. For example, if the client web browser navigated to the URL "" would return "/path/resource?a=1&b=2".

This is the raw query string, exactly as the client sent it. Any '%xx' character sequences are still present in this version of the string.

getQueryParam (name)httpreq.h[100]
Get a parameter from the query string. This parses the query string and returns the parameter with the specified name, if present, or nil if not. This does the same parsing as parseQuery(), but it returns just the specified parameter value rather than building a lookup table with all of the parameters. This is more efficient than parseQuery() if you're just looking up one or two parameters, since it avoids building the whole table, but it's less efficient if you're looking up many parameters because this routine has to re-parse the query string on each call.

getServer ( )httpreq.h[35]
Get the HTTPServer object. This is the server that received the network request that 'self' represents.

getVerb ( )httpreq.h[48]
Get the verb. This is the HTTP verb that the client sent with the request to indicate what action to perform. The standard HTTP verbs are GET (this is the most common request type; it simply retrieves a resource from the server), POST (used to submit form data), OPTIONS, HEAD, PUT, DELETE, TRACE, CONNECT, and PATCH. If the client sends a non-standard verb, the system simply passes it through to the request, allowing you to write a custom server for a custom client; however, this isn't recommended, since proxies and firewalls often block what they consider ill-formed requests.

parseQuery ( )httpreq.h[87]
Parse the query string. This returns a LookupTable containing the parsed elements of the query. The element table[1] is the base resource name: this is the part of the query string up to the first '?', if any, or the entire query string if there are no parameters. If there are any parameters, they're entered in the table under the parameter names as keys. For example, for this query string:

we'd generate this table:

table[1] = '/path/resource'
table['a'] = 'one'
table['b'] = 'two'
table['c'] = 'three'
table['d'] = ''

For all of the strings (including the base resource name, the parameter names, and the parameter values), any '%xx' sequences are converted into the corresponding character values. For example, if the base resource name is '/C%C3%A1fe', the resource name will be parsed to 'Cfe' (%C3%A1 is the percent-encoded representation of the UTF-8 character small A with acute accent).

sendReply (body, contentType?, status?, headers?)httpreq.h[289]
Send the reply to the request.

'body' is the content of the reply. For example, for a request for an HTML page, this would contain the HTML text of the page. This can be a string or StringBuffer, in which case the string is sent as UTF-8 data; a ByteArray, in which case the raw bytes of the byte array are sent exactly as given; an integer, in which case the body will be a default HTML page automatically generated for the corresponding HTML status code; a File object opened with read access, in which case the contents of the file will be sent; or nil, in which case no message body is sent at all (just the status code and the headers). A file open in text mode will be sent as text, and raw mode will be sent as binary. "Data" mode isn't allowed. Note that if a file is used, this will send the entire contents of the file, regardless of the current seek position, and the routine will have the side effect of setting the file's seek position to end of file upon return.

'contentType' is an optional string giving the MIME type of the reply body. This is used to generate a Content-type header in the reply. If this is omitted, the server generates a default MIME type according to the type of the 'body' argument. If 'body' is given as a string, and the string starts with <HTML (ignoring any initial whitespace), the default content type is "text/html"; if the string starts with <?XML, the default type is "text/xml"; otherwise the default type is "text/plain". If 'body' is a ByteArray, the server checks the first few bytes for signature strings for a few known types (JPEG, PNG, MP3, Ogg, MIDI), and uses the corresponding MIME type if matched; otherwise the MIME type is set to "application/octet-stream". 'contentType' is ignored when the 'body' argument is an integer (because in this case the body is known to be HTML), or hwen it's nil (since there's no body at all in this case).

When the 'body' is a string, the server automatically adds the "charset" parameter to the Content-type header with value "utf-8".

'status' is an optional HTTP status code. This can be a string, in which case it must have the standard format for an HTTP status code, which consists of a decimal status code number followed by the text description of the code: for example, "404 Not Found. This argument can alternatively be a simple integer giving one of the standard HTTP status code numbers, in which case we'll automatically add the corresponding status code text. If this argument is omitted, AND the 'body' is given as an integer, the 'body' value will be used as the status code. If this is omitted and there's a non-integer 'body' argument, we use "200 OK" as the default status code.

'headers' is an optional list of header strings. Each element of the list is a string giving one header, in the standard format "Name: Value". If this is omitted, the reply will contain only the standard headers synthesized by the server.

The server automatically generates certain headers with each reply. These should not be specified in the 'headers' argument. The standard headers are:

Content-type: per the 'contentType' argument
Content-length: length in bytes of 'body'

After sending the reply, the request is completed, and no further reply can be sent.

sendReplyAsync (body, contentType?, status?, headers?)httpreq.h[375]
Send the reply to the request asynchronously. This works like sendReply(), except that this method starts a new background thread to handle the data transfer and then immediately returns. This allows the caller to continue servicing other requests while the data transfer proceeds, which is important when sending a large file (such as a large image or audio file) as the reply body. Most browsers allow the user to continue interacting with the displayed page while images and audio files are transfered in the background, so it's likely that the browser will generate new requests during the time it takes to send a single large reply body. When you use sendReply(), the server won't be able to service any of those new requests until the reply is fully sent, so the browser will appear unresponsive for the duration of the reply data transfer. Using sendReplyAsync() allows you to service new requests immediately, without waiting for the data transfer to complete.

The parameters are the same as for sendReply(). If 'body' is a File, this function opens its own separate handle to the file, so you're free to close the File object immediately, or to continue to use it for other operations. Note that if you continue writing to the file after calling this method, it's unpredictable whether the reply data will contain the original or updated data (or a mix of new and old data), since the reply data transfer is handled in a separate thread that runs in parallel with the main program.

When the reply data transfer is completed, or if it fails, the system posts a NetEvent of type NetEvReplyDone to the network message queue. The event contains the original HTTPRequest object, to allow you to relate the event back to the request that generated the reply, and status information indicating whether or not the transfer was successful.

sendReplyCatch (reply, contType?, stat?, headers?)webui.t[122]
Send a reply, catching "socket disconnect" exceptions. In most cases, server objects will want to use this method rather than the native sendReply() so that they won't have to handle disconnect exceptions manually.

Disconnect exceptions are common with HTTP: the protocol is stateless by design, so clients can close their sockets at any time. Most modern browsers use HTTP 1.1, which allows the client to maintain an open socket indefinitely and reuse it serially for multiple requests, but browsers tend to close these reusable sockets after a few minutes of inactivity.

When this routine encounters a disconnected socket, it deletes the client session record for the request, and then otherwise ignores the error. This generally makes the client's socket management transparent to the server, since if the client is still running they'll just connect again with a new socket and retry any lost requests.

sendReplyChunk (body)httpreq.h[324]
Send a piece of a chunked reply. This can be called any number of times after calling startChunkedReply() to send the pieces of a chunked reply.

'body' can be a string or StringBuffer, in which case the string data are sent as UTF-8 text; or a ByteArray, in which case the raw bytes are sent.

setCookie (name, value)httpreq.h[223]
Set a cookie in the reply. This sets a cookie with the given name and value, both given as strings. The value string starts with the text to set for the cookie, and can be followed by additional parameters, delimited by semicolons:

expires=Fri, 31-Dec-2010 23:59:59 GMT - set expiration date - scope cookie to domain
path=/ - scope cookie within site
httponly - hide cookie from Javascript
secure - only send via https://

(These are all defined by the HTTP protocol, not by TADS. For details, refer to any HTTP reference book or web site.)

A cookie without an expiration date is a session cookie: it implicitly expires as soon as the browser application terminates. The presence of an expiration date makes the cookie persistent, meaning it's to be stored on disk until the given expiration date and should survive even after the browser is closed.

To send a cookie, you must call this BEFORE sending the reply or starting to send a chunked reply. This is a limitation of the protocol, since the cookies must be sent at the start of the reply with the headers. Calling this routine doesn't actually send anything immediately to the client, but simply stores the cookie with the pending request, to be sent with the reply.

startChunkedReply (contentType, resultCode?, headers?)httpreq.h[313]
Start a chunked reply. This sends the initial headers of a reply that will be generated in pieces. This is an alternative to sending the entire reply as a single string via sendReply(), for situations where you're generating the reply algorithmically and want to send a little bit at a time, rather than buffering up the entire reply in a string or StringBuffer to send all at once. This is particularly useful for cases where the reply will take a while to generate, since it allows the client to interpret partial data before the entire reply is completed; and for very large reply bodies, where it would consume a lot of memory to buffer the whole reply.

This routine doesn't send any initial data, but simply begins the reply process. For this reason, the content type must be specified: there's no way for the routine to infer the content type here since we don't have any body data to look at. The content type, result code, and headers arguments work as they do with sendReply().

After calling this routine, call sendReplyChunk() as many times as needed to send the pieces of the reply. After sending all of the pieces, call endChunkedReply() to finish the reply.

TADS 3 Library Manual
Generated on 5/16/2013 from TADS version 3.1.3