&kommander; New Parser Michal Rudolf
mrudolf@kdewebdev.org
Eric Laffoon
eric@kdewebdev.org
2005-2008 Michal Rudolf Eric Laffoon &FDLNotice;
New Parser Documentation The new parser was introduced in &kommander; with version 1.2, released with KDE 3.4. This document was originally released to show all the features of new parser. As of &kommander; 1.3, released with KDE 3.5.9, the new parser is now the default, except for MainWindow applications created in &Qt; Designer. Because the new parser is so much richer in ability, overcomes the limitations of nesting in the old parser and adds so many new features we strongly recommend using it. &kommander; itself will not be described here. Please refer to other documents to see what is &kommander; for, how to create dialogs and how to manipulate widgets on runtime. Old parser Here we compare the two parsers. While we advocate the new one for most purposes the old one is still supported and useful, particularly when working with other scripting languages. Old parser The old parser was in fact macro parser. Only strings beginning with @ were recognized, locally parsed and expanded. @LineEdit1.setText(@ListBox.selection) All the underlying functionality (local variables, expressions, file manipulation) had to be done in another scripting language, such as Bash. While the intent with &kommander; is to support all other scripting languages, and this is presently possible to some degree, there was a need for a fast, native scripting language that was assured to be portable. The biggest problem with the old parser is that the &kommander; specials are evaluated before the code is passed to the scripting language, making them impossible to use in loops and conditions. The developers considered bash slow and not friendly to new users, and the old parser had been initially bash calling DCOP. The paradox for &kommander; being language neutral resulted in a need to do more than just functions natively. New parser The new parser is a full parser. It parses the whole script, not just functions. As we were interested in GUI interaction, not the proliferation of scripting languages, we made compromises. As a result you should find &kommander;'s scripting to be capable for most basic tasks and natural and easy to use. There is also the Function Browser, which will help you assemble statements. The Function Browser is intended to make &kommander; accessible to complete novice programmers. It is similar to what you would find in KSpread to help you choose a function and fill in the parameters. If you want enhanced functionality found in other languages you can include them in &kommander; script objects headed with a shebang. While in these scripts the Function Browser will help you insert references to widgets. Just remember when using this functionality that the parser makes one pass for the old parser functions and one pass for your script. So if you try to change something in a widget and read it in the middle of a script you may not get what you expect. #!/usr/bin/php The following feature list is from version 1.2 local and global variables and associative arrays numerical expressions string manipulation various structure commands: if, while, for, foreach most functions from old parser direct widget manipulation many additional functions decent execution speed receive parameters from signals in script slots This list is from version 1.3 pass parameters and receive them with script execute calls return a value from a script create widgets on the fly connect signals and slots on the fly use a variable alias for a widget name simple indexed array functions directly access a widgets slots Invoking new parser To enable new parser, set useInternalParser property of the dialog to true. You can also enable new parser in a single script by putting #!kommander on the first line of the script. Also note if you are using another scripting language in a script with a shebang that &kommander; automatically enables the old parser for interacting with the dialog. #!/bin/bash echo @Self.item(0) # returns first parameter passed to script # echo $returnvalue passes back to calling script New Parser Features Types Each value is of one of three types: string, integer or double. Type conversion is automatic and chooses most appropriate type (for example, if you add double to integer, result will be double). If one of the values is string, result will be string too. Places you can get into trouble here are getting a numerical value from a widget and trying to perform a mathematical function on it. Since &kommander; uses + to concatonate two strings of text it can treat LineEdit1.text + 2 as 22 instead of 2. See the conversion functions in String functions to avoid problems. Expressions The following mathematical operators are supported: +, -, *, mod, . Standard brackets are of course supported as well. All kinds of comparisons are supported: <, >, <=, >=, ==, !=. Instead of != you can also use <>. Also, logical operators and, or, not are supported, as well as their C equivalents (&&, ||, !). For strings you can use + operator for string concatenation. Some examples of valid expressions: 2+3 -5 * (2 - 13 mod 3) "This list has " + 12 + "items." Variables Variables don't need to be declared. Once you use variable, it is considered declared. Type of a variable is recognized automatically and can be changed later. Associative arrays are supported as well. They map string keys onto values of any type. To declare such array, you can just add some element to it, for example: A["Quanta"] = "Web editor". Arrays are also handled by foreach command and array functions. Local and global variables are supported. Global variables are marked by leading underscore. So, myVar is a local variable, but _myVar is global. The same applies to arrays. a = 5 b = 2 * 5 - (a + 1) c = "[Item " + b + "]" d["MyKey"] = "MyValue" d["MyKey2"] = 5 Using variables for widgets works much as you would expect. This is useful when looping widgets into a table. for i=0 to 10 do mycombo = "ComboTable"+i createWidget(mycombo, "ComboBox", "Form1") end Comments You can use comments in &kommander; using the two traditional program language comment forms for line comments. For those users who are new to programming wondering what traditional form? see below. You can copy and paste the text below into a button or dialog initialization and see how comments behave in action. // this is a comment for one line message_info("Hello World") //traditional first program // the above comment also ignored - the messagebox is not # this is also a comment message_info("This message will show") Using the following multi-line comment will not work and will cause the rest of the widget execution to fail. /* Hi, I was supposed to be a comment None of the script after this will execute DON'T USE THIS TYPE OF COMMENT IN KOMMANDER! */ Built in Globals &kommander; has some built in globals you may find handy. _ARGS - the argument string passed to the dialog on opening _ARGCOUNT - the count of arguments passed. These can be retrieved as ARG1 to ARGn where n is the total number of args passed _KDDIR - the directory from which the dialog was run. &kommander; will default to your home directory, or a directory change if asked for it's current directory. This is useful for saving and reading files with the &kommander; file. _NAME - there is no reason to use this so don't _PID - the process id the current dialog is being run from - also available as just pid Avoid using this name for your variables! _VERSION - this is handy if you want to display the version of &kommander; that is running Passing arguments in &kommander; You can pass arguments via script parameters, signals and slots, command line parameters and DCOP. Let's look at scripts. Call your script like: result = ScriptObject1.execute("Hello World") debug(result) Inside your script you might have the following var = str_upper(Self.Item(0)) return(var) Now you will get a return in your Stderr message log of HELLO WORLD Receiving a signal connected to a script slot works the same way. Self.Item(0) is parameter one and so on. You can retrieve the count of arguments passed with ScriptObject.count. Command line parameters allow for named or unnamed arguments. Unnamed look like kmdr-executor myprog.kmdr 100 red Where you will find _ARG1 = 100 and _ARG2 = red. One quirk is passing strings with spaces as an argument means they need to be quoted. Using the dialog command complicates matters as the entire argument string must pass as one string, meaning in quotes. dialog("mydialog.kmdr", 100+" \"Hello World\"") This returns _ARG1 = 100 and _ARG2 = Hello World. Without the escaped quotes you would have _ARG2 = Hello and _ARG3 = World. Using Named Parameters is rather nice and potentially less confusing. dialog("mydialog.kmdr", "xcount=100 xquote=Hello world") And now you access those with _xcount and _xquote global variables. DCOP can be complex, which is why we recommend using the tools we develop to enable creating DCOP for remote &kommander; dialogs with something like a function browser. Here is an example DCOP call issued from a dialog opened from a parent &kommander; window. Since it knows who its parent is it can send information back while it is open and freely access all its parent's functionality with the exception of slots. Of course that can be done internally with a script which can be called externally, so in practice there is no limit to what can be done. dcop("kmdr-executor-"+parentPid, "KommanderIf", "setText(TQString,TQString)", "StatusBar8", "Hello") Let's look at this piece by piece. First of all we add parentPid to "kmdr-executor-" as we make no assumption a &kommander; window was the caller. You could use this with Quanta or KSpread or whatever. Next we are addressing KommanderIf, which is a nice interface for end users which has been cleaned up. We hope eventually as KDE moves from DCOP to DBUS on KDE4 that more applications adopt a nice interface for integration. The next parameter, "setText(TQString,TQString)" is important because it prototypes the parameters allowed. Otherwise &kommander; could not validate the call. So without a definition of the DCOP call being used you will get an error. The remaining parameters are of course what is being passed. We recommend you look at applications with kdcop to see how this works and practice dcop calls from the shell to get your syntax right. Commands Various structure commands are supported. They can be freely nested. There are also three special commands: exit, break and continue. The first one ends script execution and returns. The second exits current block (while, for or foreach and the third exits just a current step, restarting from the beginning of the loop. if Command if has following syntax: if condition then code elseif condition then code else code endif Both elseif and else parts are optional. Condition is any expression. Code is executed if condition is true. That means: non-zero for integers and double non-empty for strings if a * 2 > 7 then b = 1 elseif a < 0 then b = 2 elseif b = 0 endif while while condition do code end Condition is recalculated each time loop is executed. while i < 15 do i = i + a end for Command for has following syntax: for variable = start value to end value step expression do code end Loop is executed starting from start value and it is ended when variable's value is bigger then end value. If step part is specified, on each step variable's value is increased by given value instead of 1. foreach i = 1 to 20 step 5 do a = a + 2 * i end foreach Command foreach has following syntax: for variable in array do code end Loop is executed for each key in given array. In each step variable is assigned the next key from the array. sum = 0 foreach i in myArray do sum = sum + myArray[i] end Functions Most old parser functions are supported by new parser. Also, some new functions were added. String functions String functions are the same as in old parser, the only difference is that their names are preceeded by str_ instead of @String. str_length(string) - returns length of string str_contains(string, text) - returns 1 if string contains text str_find(string, text, start) - returns position of the first occurrence of text in string; optional start specifies start of the search str_find(string, text, start) - returns position of the last occurrence of text in string; optional start specifies start of the search str_left(string, count) - returns first count characters of string str_right(string, count) - returns last count characters of string str_right(string, start, count) - returns substring of string starting from start and containing count characters (or everything to the end of the string if last parameter is not specified) str_remove(string, text) - returns string with all substrings equal to text removed str_replace(string, text, text2) - returns string with all substrings equal to text replaced with text2 str_lower(string) - returns string converted to lowercase str_upper(string) - returns string converted to uppercase str_section(string, separator, start, end) - returns substring containing appropriate sections of string determined by separator; if no end is given, single start section is returned str_args(string, ...) - returns string with %1, %2, %3 replaced with following parameters. str_isnumber(string) - returns 1 if string is a valid number str_isempty(string) - returns 1 if string is empty str_toint(string, default) - returns string converted to integer; if conversion is not possible, optional default value is returned str_todouble(string, default) - returns string converted to double; if conversion is not possible, optional default value is returned &kommander; functions Most &kommander; functions are supported; some (such as expr) were obsoleted by new parser and are not available. debug(string, ...) - writes all parameters on stderr echo(string, ...) - writes all parameters on stdout dcop(string, ...) - calls DCOP function exec(string, shell) - executes external program (using optional shell); block the execution of the current dialog until the program passed as the parameter exits; returns output of that program i18n(string) - marks string for future translation env(string) - returns a value of environmental variable readSetting(key, default) - returns a value stored in config file with given key; if there is no such value default is returned writeSetting(key, value) - writes pair key and value in config file New in &kommander; 1.3 execBackground(string, shell) - executes external program (using optional shell) in the background, without blocking the current dialog; contrary to the above exec function, it will not return the output of the program. return(value) - returns a value to the calling object (script, button...) createWidget(widgetname, widgettype, parent) - creates a new widget. You can then place it in a table or toolbox, for example and use mywidget.show(true) to make it visible. If you are putting an new widget on the form you need to consider layout issues. &kommander; will not create layouts on the fly or edit pixel by pixel positioning (in most cases). This is confusing even in C++ development. We recommend you use a groupbox and do a layout in the dialog for best control. connect(sender, signal, receiver, slot) - connect a widget signal to a widget slot. See the connection dialog and select similar widgets for possibilities. If for instance a signal looks like looks like execute(const TQString&) that is exactly what must be in quotes there. disconnect(sender, signal, receiver, slot) - undo the connection as listed above. Again, exact syntax is essential. widgetExists(widgetname) - remember you can use a variable name to reference a widget now. Use this when accessing created widgets to insure they are there. Calling a non-existant widget obviously will throw an error. Array functions Most array functions are supported; some (such as value) were obsoleted by new parser and are not available. The only difference is that their names are preceeded by array_ instead of @Array. Due to parser limitation, name of array has to be specified as string now; for example array_count("MyArray"). array_clear(array) - removes all elements from array array_count(array) - returns number of elements in array array_keys(array) - returns string containing EOL-separated keys of array - note that if you had imported a scalar (keys without values, see below for an example) into an array with &kommander; you would not be able to access it with array_values("myarray") as you might think (since it seems to only have values) but would instead need to use array_keys("myarray"). You might find a better choice for this is to use the new indexed arrays described below. array_values(array) - returns string containing EOL-separated values of array array_tostring(array) - returns string containing whole array as EOL-separated pairs containing key and value separated with TAB character array_fromstring(array, string) - reads array from string (usually provided by array_tostring function) array_remove(array, key) - removes item with key key from array Here is an example for array manipulation: array_fromstring("myArray", "1\tA\nsecond\tB\n3\tC") foreach key in myArray do debug("myArray[" + key + "]= " + myArray[key]) end This will print out the following to the stderr. It is visible that there is no guarantee about the order of array elements, as well that the keys are strings, not numbers. myArray[1]= A myArray[3]= C myArray[second]= B Another example for keyless arrays: array_fromstring("myArray", "A\nB\nC") foreach key in myArray do debug(key) end debug("Array elements:\n" + array_keys("myArray")) This results in: A B C Array elements: A B C New in &kommander; 1.3 array_indexedFromString(array, string, separator) - this compensates for &kommander; not having indexed arrays. it creates an array with a zero based sequential index. Remember to use quotes on the array name and any strings not represented by a variable. The separator argument is optional and defaults to "\t" [TAB] which is used to separate fields reading and writing tables, arrays or detail widgets. Remember this array index does not enforce any rules on its self. It is just like you created it with a for loop, just more convenient. array_indexedInsertElements(array, key, string, separator) - this function is part of the indexed array suite and enables you to insert elements in your array while maintaining an index that is sequential, contiguous and unique. Set the index key to start at and the text string and how it is separated. The elements will be added shifting all the index numbers after by the number added. array_indexedRemoveElements(array, key start, number) - this enables you to remove elements from an indexed array and avoid gaps in your index. Specify the key to start at and optionally how many to remove. The default is one. You will end up with a re-indexed array less the removed elements. array_indexedToString(array, separator) - this enables you to convert your indexed array back into a string, particularly useful for detail widgets. For instance if you are displaying a database query result in TreeWidget1 and it has six columns you can use TreeWidget1.selection to get the selected row. it will be separated by tabs and you could look at a the fifth element by using str_section(TreeWidget1.selection, "\t", 4) (remember it is zero based). That's nice for reading a value, but if you want to change it you can see you have a lot more work to do. After you split that string you have to reassemble with val1+"\t"+val2... Using indexed arrays you could edit that fifth element like so... idx = TreeWidget1.currentItem array_indexedFromString("z", TreeWidget1.selection) z[4] = "new value" TreeWidget1.removeItem(idx) TreeWidget1.insertItem(array_indexedToString("z"), idx) Note that only two short lines were added to accomplish this! This was very welcome for database use. File functions All file functions are supported, the only difference is that their names are preceeded by file_ instead of @File. file_read(name) - returns content of file name file_write(name, ...) - writes all arguments to file name file_append(name, ...) - appends all arguments to file name Input functions These functions show some dialog allowing user to enter some value. They are accessible in the old parser using @Input.. For most functions all parameters are optional, exception is input_text which requires 2 parameters and input_value which requires 5 parameters. input_color(caption, default) - returns color in #RRGGBB format input_text(caption, label, default) - returns text entered by user input_value(caption, label, default, min, max, step) - returns value entered by user input_directory(startdir, filter, caption) - returns directory selected by user input_openfile(caption, label, default) - returns existing file entered by user input_savefile(caption, label, default) - returns file entered by user (if file exists, confirmation will be required) input_openfiles(caption, label, default) - returns string of EOL-separated existing files entered by user Message functions These functions show some message for user or ask user to confirm some action. In the old parser use @Message. instead. message_info(text, caption) - shows information text message_error(text, caption) - shows error text message_warning(text, caption, button1, button2, button3) - shows question with warning and up to three buttons; number of chosen button is returned; if no button names are specified, Yes and No will be displayed message_question(text, caption, button1, button2, button3) - shows question and up to three buttons; number of chosen button is returned; if no button names are specified, Yes and No will be displayed