***************************************************** Using commands.wrap to parse your command's arguments ***************************************************** This document illustrates how to use the new 'wrap' function present in Supybot 0.80 to handle argument parsing and validation for your plugin's commands. Introduction ============ To plugin developers for older (pre-0.80) versions of Supybot, one of the more annoying aspects of writing commands was handling the arguments that were passed in. In fact, many commands often had to duplicate parsing and verification code, resulting in lots of duplicated code for not a whole lot of action. So, instead of forcing plugin writers to come up with their own ways of cleaning it up, we wrote up the wrap function to handle all of it. It allows a much simpler and more flexible way of checking things than before and it doesn't require that you know the bot internals to do things like check and see if a user exists, or check if a command name exists and whatnot. If you are a plugin author this document is absolutely required reading, as it will massively ease the task of writing commands. Using Wrap ========== First off, to get the wrap function, it is recommended (strongly) that you use the following import line:: from supybot.commands import * This will allow you to access the wrap command (and it allows you to do it without the commands prefix). Note that this line is added to the imports of plugin templates generated by the supybot-plugin-create script. Let's write a quickie command that uses wrap to get a feel for how it makes our lives better. Let's write a command that repeats a string of text a given number of times. So you could say "repeat 3 foo" and it would say "foofoofoo". Not a very useful command, but it will serve our purpose just fine. Here's how it would be done without wrap:: def repeat(self, irc, msg, args): """ Repeats times. """ (num, text) = privmsg.getArgs(args, required=2) try: num = int(num) except ValueError: raise callbacks.ArgumentError irc.reply(num * text) Note that all of the argument validation and parsing takes up 5 of the 6 lines (and you should have seen it before we had privmsg.getArgs!). Now, here's what our command will look like with wrap applied:: def repeat(self, irc, msg, args, num, text): """ Repeats times. """ irc.reply(text * num) repeat = wrap(repeat, ['int', 'text']) Pretty short, eh? With wrap all of the argument parsing and validation is handled for us and we get the arguments we want, formatted how we want them, and converted into whatever types we want them to be - all in one simple function call that is used to wrap the function! So now the code inside each command really deals with how to execute the command and not how to deal with the input. So, now that you see the benefits of wrap, let's figure out what stuff we have to do to use it. Syntax Changes ============== There are two syntax changes to the old style that are implemented. First, the definition of the command function must be changed. The basic syntax for the new definition is:: def commandname(self, irc, msg, args, , , ...): Where arg1 and arg2 (up through as many as you want) are the variables that will store the parsed arguments. "Now where do these parsed arguments come from?" you ask. Well, that's where the second syntax change comes in. The second syntax change is the actual use of the wrap function itself to decorate our command names. The basic decoration syntax is:: commandname = wrap(commandname, [converter1, converter2, ...]) .. note:: This should go on the line immediately following the body of the command's definition, so it can easily be located (and it obviously must go after the command's definition so that commandname is defined). Each of the converters in the above listing should be one of the converters in commands.py (I will describe each of them in detail later.) The converters are applied in order to the arguments given to the command, generally taking arguments off of the front of the argument list as they go. Note that each of the arguments is actually a string containing the NAME of the converter to use and not a reference to the actual converter itself. This way we can have converters with names like int and not have to worry about polluting the builtin namespace by overriding the builtin int. As you will find out when you look through the list of converters below, some of the converters actually take arguments. The syntax for supplying them (since we aren't actually calling the converters, but simply specifying them), is to wrap the converter name and args list into a tuple. For example:: commandname = wrap(commandname, [(converterWithArgs, arg1, arg2), converterWithoutArgs1, converterWithoutArgs2]) For the most part you won't need to use an argument with the converters you use either because the defaults are satisfactory or because it doesn't even take any. Customizing Wrap ================ Converters alone are a pretty powerful tool, but for even more advanced (yet simpler!) argument handling you may want to use contexts. Contexts describe how the converters are applied to the arguments, while the converters themselves do the actual parsing and validation. For example, one of the contexts is "optional". By using this context, you're saying that a given argument is not required, and if the supplied converter doesn't find anything it likes, we should use some default. Yet another example is the "reverse" context. This context tells the supplied converter to look at the last argument and work backwards instead of the normal first-to-last way of looking at arguments. So, that should give you a feel for the role that contexts play. They are not by any means necessary to use wrap. All of the stuff we've done to this point will work as-is. However, contexts let you do some very powerful things in very easy ways, and are a good thing to know how to use. Now, how do you use them? Well, they are in the global namespace of src/commands.py, so your previous import line will import them all; you can call them just as you call wrap. In fact, the way you use them is you simply call the context function you want to use, with the converter (and its arguments) as arguments. It's quite simple. Here's an example:: commandname = wrap(commandname, [optional('int'), many('something')]) In this example, our command is looking for an optional integer argument first. Then, after that, any number of arguments which can be anything (as long as they are something, of course). Do note, however, that the type of the arguments that are returned can be changed if you apply a context to it. So, optional("int") may very well return None as well as something that passes the "int" converter, because after all it's an optional argument and if it is None, that signifies that nothing was there. Also, for another example, many("something") doesn't return the same thing that just "something" would return, but rather a list of "something"s. Converter List ============== Below is a list of all the available converters to use with wrap. If the converter accepts any arguments, they are listed after it and if they are optional, the default value is shown. id, kind="integer" Returns something that looks like an integer ID number. Takes an optional "kind" argument for you to state what kind of ID you are looking for, though this doesn't affect the integrity-checking. Basically requires that the argument be an integer, does no other integrity-checking, and provides a nice error message with the kind in it. ip Checks and makes sure the argument looks like a valid IP and then returns it. int, type="integer", p=None Gets an integer. The "type" text can be used to customize the error message received when the argument is not an integer. "p" is an optional predicate to test the integer with. If p(i) fails (where i is the integer arg parsed out of the argument string), the arg will not be accepted. index Basically ("int", "index"), but with a twist. This will take a 1-based index and turn it into a 0-based index (which is more useful in code). It doesn't transform 0, and it maintains negative indices as is (note that it does allow them!). color Accepts arguments that describe a text color code (e.g., "black", "light blue") and returns the mIRC color code for that color. (Note that many other IRC clients support the mIRC color code scheme, not just mIRC) now Simply returns the current timestamp as an arg, does not reference or modify the argument list. url Checks for a valid URL. httpUrl Checks for a valid HTTP URL. long, type="long" Basically the same as int minus the predicate, except that it converts the argument to a long integer regardless of the size of the int. float, type="floating point number" Basically the same as int minus the predicate, except that it converts the argument to a float. nonInt, type="non-integer value" Accepts everything but integers, and returns them unchanged. The "type" value, as always, can be used to customize the error message that is displayed upon failure. positiveInt Accepts only positive integers. nonNegativeInt Accepts only non-negative integers. letter Looks for a single letter. (Technically, it looks for any one-element sequence). haveOp, action="do that" Simply requires that the bot have ops in the channel that the command is called in. The action parameter completes the error message: "I need to be opped to ...". expiry Takes a number of seconds and adds it to the current time to create an expiration timestamp. literal, literals, errmsg=None Takes a required sequence or string (literals) and any argument that uniquely matches the starting substring of one of the literals is transformed into the full literal. For example, with ``("literal", ("bar", "baz", "qux"))``, you'd get "bar" for "bar", "baz" for "baz", and "qux" for any of "q", "qu", or "qux". "b" and "ba" would raise errors because they don't uniquely identify one of the literals in the list. You can override errmsg to provide a specific (full) error message, otherwise the default argument error message is displayed. to Returns the string "to" if the arg is any form of "to" (case-insensitive). nick Checks that the arg is a valid nick on the current IRC server. seenNick Checks that the arg is a nick that the bot has seen (NOTE: this is limited by the size of the history buffer that the bot has). channel Gets a channel to use the command in. If the channel isn't supplied, uses the channel the message was sent in. If using a different channel, does sanity-checking to make sure the channel exists on the current IRC network. inChannel Requires that the command be called from within any channel that the bot is currently in or with one of those channels used as an argument to the command. onlyInChannel Requires that the command be called from within any channel that the bot is currently in. nickInChannel Requires that the argument be a nick that is in the current channel, and returns that nick. networkIrc, errorIfNoMatch=False Returns the IRC object of the specified IRC network. If one isn't specified, the IRC object of the IRC network the command was called on is returned. callerInGivenChannel Takes the given argument as a channel and makes sure that the caller is in that channel. plugin, require=True Returns the plugin specified by the arg or None. If require is True, an error is raised if the plugin cannot be retrieved. boolean Converts the text string to a boolean value. Acceptable true values are: "1", "true", "on", "enable", or "enabled" (case-insensitive). Acceptable false values are: "0", false", "off", "disable", or "disabled" (case-insensitive). lowered Returns the argument lowered (NOTE: it is lowered according to IRC conventions, which does strange mapping with some punctuation characters). anything Returns anything as is. something, errorMsg=None, p=None Takes anything but the empty string. errorMsg can be used to customize the error message. p is any predicate function that can be used to test the validity of the input. filename Used to get a filename argument. commandName Returns the canonical command name version of the given string (ie, the string is lowercased and dashes and underscores are removed). text Takes the rest of the arguments as one big string. Note that this differs from the "anything" context in that it clobbers the arg string when it's done. Using any converters after this is most likely incorrect. glob Gets a glob string. Basically, if there are no wildcards (``*``, ``?``) in the argument, returns ``*string*``, making a glob string that matches anything containing the given argument. somethingWithoutSpaces Same as something, only with the exception of disallowing spaces of course. capability Used to retrieve an argument that describes a capability. channelDb Sets the channel appropriately in order to get to the databases for that channel (handles whether or not a given channel uses channel-specific databases and whatnot). hostmask Returns the hostmask of any provided nick or hostmask argument. banmask Returns a generic banmask of the provided nick or hostmask argument. user Requires that the caller be a registered user. matches, regexp, errmsg Searches the args with the given regexp and returns the matches. If no match is found, errmsg is given. public Requires that the command be sent in a channel instead of a private message. private Requires that the command be sent in a private message instead of a channel. otherUser Returns the user specified by the username or hostmask in the argument. regexpMatcher Gets a matching regexp argument (m// or //). validChannel Gets a channel argument once it makes sure it's a valid channel. regexpReplacer Gets a replacing regexp argument (s//). owner Requires that the command caller has the "owner" capability. admin Requires that the command caller has the "admin" capability. checkCapability, capability Checks to make sure that the caller has the specified capability. checkChannelCapability, capability Checks to make sure that the caller has the specified capability on the channel the command is called in. Contexts List ============= What contexts are available for me to use? The list of available contexts is below. Unless specified otherwise, it can be assumed that the type returned by the context itself matches the type of the converter it is applied to. any Looks for any number of arguments matching the supplied converter. Will return a sequence of converted arguments or None. many Looks for multiple arguments matching the supplied converter. Expects at least one to work, otherwise it will fail. Will return the sequence of converted arguments. optional Look for an argument that satisfies the supplied converter, but if it's not the type I'm expecting or there are no arguments for us to check, then use the default value. Will return the converted argument as is or None. additional Look for an argument that satisfies the supplied converter, making sure that it's the right type. If there aren't any arguments to check, then use the default value. Will return the converted argument as is or None. rest Treat the rest of the arguments as one big string, and then convert. If the conversion is unsuccessful, restores the arguments. getopts Handles --option style arguments. Each option should be a key in a dictionary that maps to the name of the converter that is to be used on that argument. To make the option take no argument, use "" as the converter name in the dictionary. For no conversion, use None as the converter name in the dictionary. first Tries each of the supplied converters in order and returns the result of the first successfully applied converter. reverse Reverse the argument list, apply the converters, and then reverse the argument list back. commalist Looks for a comma separated list of arguments that match the supplied converter. Returns a list of the successfully converted arguments. If any of the arguments fail, this whole context fails. Final Word ========== Now that you know how to use wrap, and you have a list of converters and contexts you can use, your task of writing clean, simple, and safe plugin code should become much easier. Enjoy!