Re: [PEPr] Proposal for HTTP::HTTP_Spider

On 5 Mar 2004 at 13:15, Matthew Feinberg wrote: > Is it okay to just return false instead of a pear error? Is it > preferred to return a pear error even if you only need a boolean so > that future development is easier, e.g. you've added new logic to a > function which requires the the addition of more fine grained errors. > (Sorry, I'm new to PEAR development) Depends on the way you implement your API. Generally I'd say it's okay to return false from a simple function. But please note that returning a PEAR_Error at a later revision of your package would mean a BC break (backward-compatibility). So if you just don't throw a PEAR_Error *now* but will do in the next revision, maybe already implement at least a simply error. > > >* What purpose does the empty constructor serve? > Again, none. I looked at some other classes to try to understand how > some of the PEAR coding standards are used in practice. The class I > looked at, PEAR_PackageFileManager, has an empty constructor. I'll > remove mine. PEAR_PackafeFileManager just has it for documentation purpose. Have a look at the comments above: http://cvs.php.net/co.php/pear/PEAR_PackageFileManager/PackageFileMana ger.php?r=1.34 This wouldn't have been possible without an empty constructur. [...] In general I like your proposal. But as said, you should try to PEARify it more, use existing PEAR-packages instead of reinventing the wheel etc. where apropriate. Then feel free to ask for comments again here on pear-dev. Thank you for your decision to contribute to PEAR. Regards, Stefan