As an example, Image.load could not load from URL:s since pike 7.4 or so.
The reason being that Nilsson wanted to load the HTTP module dynamically, but did not know about the difference between -> and [], with regard to dirnode modules.
Personally I thought that the support for loading from URL:s was simply removed for some other reason, and never investigated more closely.
While we are at that: I also thought it was removed for other reasons and was pretty happy about it. I'd like to deautomize the image-URL-loading. It's a nonobvious feature that could potentially halt a program for at least several seconds or add undesired "features" to a program. Something like Image.load("string", ALLOW_URLS);. URI-objects could be automaticly loaded.
Example: Site lets users choose avatar by entering the image name* that is then loaded for rescaling by Image.load.
* Probably a select-list, but it boils down to sending a string. While this is not exactly good practise, I have no problem imagining it being done.
Comments?
I think it is enough to just document it properly. The point with Image.load is to add a lot of "helpful" stuff on top of Image._decode(Stdio.read_file()). It launces external processes to decompress archives as an example.
It is not used in the Roxen tree, if that is what you were afraid of.
pike-devel@lists.lysator.liu.se
-
Martin Nilsson (DivX Inc.) ＠ Pike (-) developers forum -
Per Hedbor () ＠ Pike (-) developers forum
-
Peter Bortas ＠ Pike developers forum