<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> </head> <body text="#000000" bgcolor="#FFFFFF"> <div class="moz-cite-prefix">Hi Juan, I had a closer look at your implementation. It is less code and looks simpler, but it uses offsets and indexes, therefore I find it harder to understand. It has the following drawbacks/missing features: <ol> <li>It is not possible to pass options to scripts, they are are all consumed by the parsing code and then ether honored or ignored.</li> <li>In order to implement <tt>#lastCommandLineArguments</tt> properly, you have to look ahead for a <tt>--</tt>. Of course before parsing, the arguments after <tt>--</tt> can be stored in an instance variable </li> </ol> I am against implementing feature 1, because it implies once we add new options understood by Cuis' image, the options passed to the script would change in a rather surprising way. The attached changes set implements feature 2. It only adds an instance variable and about 7 lines of code. With it you can do something like: --- snip --- <tt>squeak -vm-display-null -headless Cuis6.0-5632cli2.image -d "Smalltalk scriptingArguments do: [ :arg | StdIOWriteStream stdout nextPutAll: arg; space]. StdIOWriteStream stdout newLine" a b c dee -d "Smalltalk scriptingArguments do: [ :arg | StdIOWriteStream stdout nextPutAll: arg; space]. StdIOWriteStream stdout newLine" z y x why -q -- 1 2 3 -o</tt> --- snap --- and get the following output: --- snip --- <tt>a b c dee 1 2 3 -o </tt><tt> </tt><tt>z y x why 1 2 3 -o </tt> --- snap --- I suppose that should support all sensible uses cases. I also presume, that all requirements of Dr. Geo are fulfilled. HTH and Best Regards, Gerald On 1/13/23 8:21 PM, Juan Vuletich wrote: </div> <blockquote type="cite" cite="mid:63C1AF4F.3000004@cuis.st">Hi Gerald, Please check the attach. Besides being simpler and hopefully easier to understand, it allows each script to have its own set of arguments. Perhaps we could add another getter to SystemDictionary, something like #lastCommandLineArguments or such, to fetch what's after '--'. Thanks, On 1/13/2023 1:18 PM, Gerald Klix via Cuis-dev wrote: <blockquote type="cite">I never received this message, therefore I am reseinding it: Hi Juan, let me answer questions one at time (even those that were not asked:) -q) If want to write small Cuis scripts, like the prototypical “Hello World” program, you would write something like this: squeak -vm-display-null -headless Cuis6.0-5626 -d "StdIOWriteStream stdout nextPutAll: 'Hello World!'; newLine. Smalltalk quit" If you want to dump all unknown contributors to a file; it would look like this: squeak -vm-display-null -headless Cuis6.0-5626 -d "Smalltalk unknownContributors printOn: StdIOWriteStream stdout. Smalltalk quit" > unknownContributors.txt Of course all these one-liners will end with `Smalltalk quit`, therefore I came up with a shorter way to achieve the same effect: squeak -vm-display-null -headless Cuis6.0-5626 -d "Smalltalk unknownContributors printOn: StdIOWriteStream stdout" -q > unknownContributors.txt -c) is completely useless, because the old implementation only accepts one `-s` option. `-s` consumes and copies all the remaining arguments to `startUpScriptArguments`, terminating the loop,I did not catch that detail. I will fix this in the next version of the change. --) Posix requires a program to stop option parsing once it encounters a double dash. To give an example that proves the necessity of this behavior, just try this in your Mac's terminal: echo Hehehe > -f Than try to delete the file named “-f” using `rm -f`. This does not work, you have to use `rm -- -f`. You can also check git, grep, ls or any other little Unix utility . Just check: echo Hehehe > -r ls -r vs. ls -- -r (I leave it as an exercise to the gentle reader to get rid of that '-r' file.) Therefore handling `--` should be done correctly, because it avoids unnecessary surprises. But it also makes it easy to pass arguments to code package (classes) that are installed with "-r". This part of Haver's installation package could be written in a less hacky way: <a class="moz-txt-link-freetext" href="https://hg.sr.ht/~cy-de-fect/HaverOnCuis/browse/haver/Haverize.pck.st?rev=tip#L131">https://hg.sr.ht/~cy-de-fect/HaverOnCuis/browse/haver/Haverize.pck.st?rev=tip#L131</a> The same is true for the `-d` option, you can implement `echo` with Cuis like this: squeak -vm-display-null -headless Cuis6.0-5626 -d "Smalltalk startUpScriptArguments do: [ :arg | StdIOWriteStream stdout nextPutAll: arg; space]. StdIOWriteStream stdout newLine" -q -- -n 1 2 3 a b dee compare to echo -- -n 1 2 3 a b dee Of course you want '--' handled correctly in your Cuis scripts, there is no use in every author implementing it again, when she needs it. #copyStartupScriptArgumentsFrom:) Simply creates a fresh copy of the arguments after `--` for each script, code file, package or piece of code passed on the command line. There is a reason for recreating the array for each script, code file etc.: I never like the Unix's getopt(3) behavior. «By default, getopt() permutes the contents of argv as it scans, so that eventually all the nonoptions are at the end.» (<a class="moz-txt-link-freetext" href="https://www.man7.org/linux/man-pages/man3/getopt.3.html">https://www.man7.org/linux/man-pages/man3/getopt.3.html</a>) Option parsing and program should no mess with the command line arguments, this creates all sorts of nasty side effects. I hope I made my point about "end of options" and "--" clear. Please give some time to come up with a simpler version of the change. HTH, Gerald On 1/12/23 3:59 PM, Juan Vuletich wrote: <blockquote type="cite">Hi Gerald, I agree that "do not ignore exceptions" doesn't need an argument. Then, maybe -e should never take an argument. I also like -h and -v. Not sure about the value of -q, but I wouldn't mind including it. But can you explain -- and #copyStartupScriptArgumentsFrom: ? Can you give an example of use? I don't get it. Thanks, On 1/11/2023 6:53 PM, Gerald Klix via Cuis-dev wrote: <blockquote type="cite">Hi guys, thanks a lot for the encouraging feedback. It motivated me to to improve the exception ignoring stuff a bit; I just wanted '-e' to work without an argument. Then I noticed, that we can't handle options without arguments. Therefore I added, em well, some other options to the command line parsing code. I hope I haven't overdone it. You can now say `squeak Cuis.image -h` and get this output: --- snip -- Command line options: -h Display this help message and quit -v Display the system information, containg the version and quit -q Just quit -e <exception> Evaluate exception and ignore it -e Set the exception to ignore to nil, i.e. do not ignore exceptions -r <feature> Require <feature> -d <code> Evaluate <code> -l <codeFile> Load e.g. file in <codeFile> -s <scriptFile> Evaluate code ein <scriptFile> -c Clear the array of start up script arguments -- end of options; stop processing options Arguments for scripts, code files, packages (and <code>): -s and -- store the remaining arguments in 'Smalltalk startUpScriptArguments' -c clears the aforementioned array Packages (-f), code files (-l) and code (-d) gets a fresh copy of the array if -- is present Multiple options: Options -e, -r, -d, -l, -s and -c can be passed multiple times --- snap --- Please review the change set and give me feedback. There might be some incompatibilities. HTH and Best Regards, Gerald On 1/11/23 2:33 PM, Dr. Nicola Mingotti wrote: <blockquote type="cite">nice ! Nicola On 1/8/23 17:21, Gerald Klix via Cuis-dev wrote: <blockquote type="cite">Hi all, Hi Juan, I was bit annoyed by the impossibility to debug scripts and packages run or installed from the command line. Therefore I added an option, ``-e <argument>``, that lets you set the exception class to handle before executing a script. With the attached change set, you can now write: squeak Cuis.image -e nil -s myScript.st If ``myScript.st`` contains an error, a Debugger will pop up. HTH and Best Regards, Gerald </blockquote> </blockquote> </blockquote> </blockquote> </blockquote> </blockquote> </body> </html>