Blocks from strings

Nevin Pratt nevin at
Thu Jan 1 19:51:26 UTC 2004

Nevin Pratt wrote:

> Avi Bryant wrote:
>> <snip>
>> Although I'm very interested in macros and how or why they might be 
>> integrated with Smalltalk, I've lost the thread of how this relates 
>> back to the original posting.  Nevin, can you tell us what you were 
>> actually trying to do by converting strings into blocks, so that I 
>> can try to understand if macros would help?
>> Avi
> Yes, but it will take me a bit to explain it.  It is in conjunction 
> with Seaside and search engine spiders.
> First, if you remember a thread on the Seaside list from back in March 
> 2003, titled "[Seaside] anchors behaving like buttons?", you'll 
> remember that I have code that allows the web user to invoke an 
> arbitrary method on a component.  For example...
> ...will invoke the #viewcart method of the 'index' component, which 
> would then take them to the shopping cart.
> This scheme allows me to email somebody a URL that will allow them to 
> enter the Seaside app at a relatively arbitrary point/page.  Once they 
> enter the app, though, it is thereafter tracked like any normal 
> Seaside app, and thereafter has the normal Seaside URL's.
> How the above scheme works is that I actually only have one component 
> (the "index", or home page component), but it has an instance variable 
> that can have different components assigned to it to control the 
> actual contents of the current webpage being displayed.  The "index" 
> component I also refer to as the "top" component.
> I also have a subclass of WASession (which I call BBSession), that 
> provides a method for accessing the "top" component (via #topComponent).
> Thus far, I have the interesting situation where I can now call any 
> arbitrary method of the top component from anywhere in the app.  For 
> example, suppose the app is on a page where the user clicks a button 
> which invokes a method of that page, and within that method I have 
> "self session topComponent viewCart".  The result is that they can 
> click the button and have the shopping cart shown to them.  Yes, I 
> know there are many other alternate ways to accomplish this same 
> thing, but bear with me...
> Next, I have added "WARenderer>>imageWithSource: srcUrl action: 
> actionBlock ", per an earlier message of yours a few months ago.  I 
> have also added "WARenderer>>imageWithSource: srcUrl link: aLink" per 
> your earlier message.  I commonly have code like this:
>   html imageWithSource: aShoppingCartJPEGLink
>           action: [self session topComponent viewcart]
> (of course, the "viewcart" message in the block above might be some 
> other method that the top component understands-- I am just using 
> #viewcart as an example).
> Now, the idea is, if the WARenderer knows it is *not* servicing a 
> spider, I want it to invoke #imageWithSource:action:, as shown above.  
> But if it is servicing a spider, I want it to automatically do this 
> instead:
>   html imageWithSource: aShoppingCartJPEGLink
>            link: ""
> This approach results in spiders automatically seeing a static site 
> with static URL's, and everybody else seeing a normal Seaside app.  A 
> way to do that would be to create a method somewhat like this:
>   html imageWithSource: aShoppingCartJPEGLink
>            method: #viewcart
> ...and then let the #imageWithSource:method: code either construct a 
> static URL or a block, and then it calls one of the earlier two methods.
> Then my code would always use #imageWithSource:method:.  All spiders 
> would then automatically see what looks to the spider like a static 
> site with normal URL's, and everybody else would automatically see a 
> normal Seaside site.  Well, almost...
> To finish the job of spiders seeing static pages, I have a page cache 
> similar to what Cees did with Janus.  So, for example, if my Seaside 
> app sees this URL coming in:
> and if the requester is a spider, it checks the page cache for the 
> #viewcart page, and if it is present, it feeds it back to the spider.  
> If it is not present, it generates it, caches it, and feeds it to the 
> spider.
> But for everybody else, it is just a normal Seaside app.
> Janus comes pretty close to this identical thing, too.
> Nevin

To see this whole thing in action, first go here:

This will make the system think you are a spider.

Now just hit the site normally:

You will see that all the Seaside gobble-de-goop is now missing from all 
URL's and all pages.

Go ahead and hover the mouse over any anchor, or any "link" image.  You 
will see that the URL that was generated for each link is also a 
static-looking link.

The site will continue to think you are a spider until one hour after 
your last access.


Nevin Pratt
Bountiful Baby
(801) 992-3137

More information about the Squeak-dev mailing list