That's a start. What kind and how could even that be improved? ~PM Date: Fri, 2 Jan 2015 15:16:00 -0500 Subject: Re: [agi] Namespace search optimization From: [email protected] To: [email protected]
A look up table?Jim Bromer On Thu, Jan 1, 2015 at 9:45 PM, Piaget Modeler via AGI <[email protected]> wrote: In the Premise language function rules are very simple. A function name follows a left parenthesis.However the name can be fully qualified <namespace>.<function> or unqualified <function> . When the name is unqualified we have to resolve the name, qualifiy it in order to select the appropriate function to execute. Each namespace has a set of imported namespaces which may contain the correctfunction definition. We only need the first namespace that has the correct function definition, we don't care if it is multiply defined. Suppose we're in the namespace User and want to access the sum function defined in the Math namespace.We have a situation like this: (using User) .: User (dependencies User) ; return the namespaces this namespace requires.: {Premise System} (namespace Math (function sum ?args (apply + ?args))).: Math (dependencies User) .: {Premise System} (Math.sum 1 2 3) ; fully qualified function call.: 6 (sum 1 2 3) ; unqualified function call Math namespace is not known to User.: [Exception :Text 'The function sum is unknown'] (require Math) ; make Math known to user.: Math (dependencies User) .: {Premise System Math} (sum 1 2 3) ; resolves sum to Math.sum.: 6 The resolution takes time. Albeit a small fraction of time. But it is cumulative. This is the specific problem. ~PM Date: Thu, 1 Jan 2015 20:39:02 -0500 Subject: Re: [agi] Namespace search optimization From: [email protected] To: [email protected] Are you referring to a problem which involves taking runtime data and searches for an appropriate function given the dynamic value? The search for the type of object that is defined in runtime is not part of the problem is it? Because a simple call where the type of the variable is given or directly implied (by uniqueness for example) should not take too much time even during runtime. I can think of two search problems that might occur in something like that. If the establishment of the appropriate set function requires some trial and error data-fitting (or function fitting) that could turn out to be inefficient. Or if the search involves a proverbial tree search then that might take some time as well. I don't see a straightforward run-time function call (to something that is like a template) as being that time consuming.Jim Bromer On Thu, Jan 1, 2015 at 8:11 PM, Piaget Modeler via AGI <[email protected]> wrote: It basically boils down to a basic speed versus extensibility tradeoff. From: [email protected] To: [email protected] Subject: RE: [agi] Namespace search optimization Date: Thu, 1 Jan 2015 08:55:44 -0800 Given 200-250 base functions in 10 packages (namespaces), if the functions are defined as language intrinsics (like if or for) then there is no need for pakckage lookup, no need to resolve the name with the namespace, so no overheard is incurred. If the functions are defined as qualified identifiers (prefixed by package name) then we need to look up any unqualified identifiers first, thereby resolving the identifier with the package, before proceeding with the evaluation. For Example, the set function in the System.KB package (namespace) sets a slot in a prototype instance to a value. The programmer can fully qualify the function call (System.KB.set ?identifier ?slot ?value) Or the programmer can reference the package and use an unqualified call (require System.KB) (set ?identifier ?slot ?value) When set is encountered we need to search the required namespaces to determinewhich set function is implied, hence we find System.KB.set and replace set withthe fully qualified name. We do this during form evaluation in the REPL,whetherthe REPL is just-in-time compiled or interpreted. If we defined set as an intrinsic then there would be no package issue, but also no modularity. We just encounter set and call the set intrinsic. So there is no overheadincurred by attempting to resolve the function name with a namespace. So, my question is, are there any known optimizations to this problem of resolving function names with packages? ~PM From: [email protected] To: [email protected] Subject: RE: [agi] Namespace search optimization Date: Thu, 1 Jan 2015 09:03:05 +0000 Dictionary lookups (on the first part of the name if the full list is big compared to RAM)? From: Piaget Modeler via AGI [[email protected]] Sent: 01 January 2015 05:53 To: AGI Subject: [agi] Namespace search optimization Are there any optimizations that can be done to look up identifiers in namespaces for either just in time compilers or interpreters ? I'm writing a REPL and namespace resolution of function identifiers takes too much time away from overall evaluation. Any ideas or thoughts? ~PM AGI | Archives | Modify Your Subscription UNIVERSITY OF CAPE TOWN This e-mail is subject to the UCT ICT policies and e-mail disclaimer published on our website at http://www.uct.ac.za/about/policies/emaildisclaimer/ or obtainable from +27 21 650 9111. This e-mail is intended only for the person(s) to whom it is addressed. If the e-mail has reached you in error, please notify the author. If you are not the intended recipient of the e-mail you may not use, disclose, copy, redirect or print the content. If this e-mail is not related to the business of UCT it is sent by the sender in the sender's individual capacity. AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription AGI | Archives | Modify Your Subscription ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
