On Mon, Jun 23, 2008 at 3:36 PM, stef <stef.dev at free.fr> wrote: > Le Thursday 19 June 2008 15:36:59 m. allan noah, vous avez ?crit : >> pretty good, but i have two comments: >> >> 1. i think calling the struct SANE_Config instead of something with >> sanei in it, makes it look too much like something that should be in >> the SANE standard, instead of the internal-use function that it is. >> >> 2. some backends seem to have dozens of options, that struct is going >> to become quite large if it is extending to hold everything. does it >> instead make more sense for the callback to be run for each line >> extracted from the file, with a flag indicating that it was an option >> instead of an attach line? >> >> allan >> > > Hello, > > here's a new take at config option parsing. The struct is now called > SANEI_Config, and holds an array of SANE_Option_Descriptor and an array of > placeholder for the values of the options.
brilliant. enables us to reuse lots of code, and keep a familiar interface. bravo. > To parse options, a backend first > creates a list of the options it wants using SANE_Option_Descriptor. great. > Then it calls sanei_configure_initialize() which creates the placeholders. > When the SANEI_Config is ready, the backend calls > sanei_configure_attach() > which opens the configuration, parse it according to the given descriptors. > Whenever a line no recognized as an option is met, the > sanei_configure_attach() calls the backend provided (*attach)(SANEI_Config *, > char *devname) attach function. This backend specific function calls the > usb/scsi/parport attach functions allready provided by sanei library. As a > parameter it has the currently parsed options in the config struct and so may > use them during the attach. A sanei_configure_get_value() helper is given to > access the values of the options. > Once parsing is done, there is a sanei_configure_dispose() function to > release allocated ressources. i would rather that the backend provided the placeholder memory if possible, for two reasons: 1. to keep all the memory allocations in one place, 2. so that the placeholders can be pre-loaded with default values by the backend. > By doing this way, configuration is parsed only with the options the > backend > needs. The existing sanei_check_value() may be reused to check if constraints > are respected. Only a couple options that need a complex syntax couldn't be > handled right now (such as 'calibrate /path/name port' from canon_pp) but > could be handled by a SANE_TYPE_STRING_LIST to be created. > > The appended patch currently only manage SANE_TYPE_STRING options and > as been > tested with a modified rts8891 backend. this sounds quite good- thank you very much for taking on the task! allan -- "The truth is an offense, but not a sin"
