edev: searching

I have just had an epithany.

It was like a bolt of lightning striking me on the head

Searching is computationally expensive.
Why do we do it so often?

I'm assuming that it's more expensive to draw up the list of near searches, than to just search for everything2.com/?node=nate; if it isn't, then I think we need to seriously sort out how we do it. A table of node name against ID would help, allowing binary searches and stuff to be done on it, blah, blah, blah.

So if we miss the node, would it be better to output 'No exact nodes found for "hypocrates". Do you want to search for this?', which for the expense of two pages instead of one may well cut down system time. And if they want to search, they can always tick the box first time.

I don't know if this will help, as it seems from my experience to be a Webserver problem (long delays) rather than Database problems.. but it might.

anotherone abuses his editorial powers: Yes, but how do we find the e2node nate without searching? If we don't have the exact node_id, a search must be performed.

the cow abuses the power of the last-person-to-edit: We have to search, that's true. But it should be much more economical to search a list sorted by node title than to search through each node_id for the correct node. True, that's more upkeep work for the server each time a fullnode is added or renamed, as well as the possibility of becoming out-of-sync with the rest of the database, but I'd think that it'd be faster than what I assume the current method is... And maybe we should include a [id:1053200|edev: searching] type link to prevent that type of searching, if people are willing?

JayBonci: I agree with you: but is this worth the lag? (Assuming that this would help, at all, anyway).

---

There is a reply to this node. Click here to view them.

A1: I think what The Cow is getting at is that when we miss a node, we give a pile of things like it. This is good and bad. Of course we have to search for "nate" but it's easier to get a dead-on hit I think than it is to get a pile of stuff " LIKE " it.

It's bad because maybe it's a little expensive. This is a good web experience, in that it is automatically searched for you.

If we searched for a "JayBonci eats foobar" via a link, we'd get a "hey!" nothing was found! and a list of applicable choices that are somewhat close to this. I think what The Cow is suggesting is that we don't give those "close" e2nodes.

...And I think that is bad. For instance, the other day I clicked on a link for The Gospel of Nicodemus in AMJ's writeup underneat Joseph of Arimathea. It had been auto-noded under Gospel of Nicodemus by dann, because that happens to be the title he chose. Without those close matches, I may not have found it, or worse yet, written up about it under a close, misspelled, or differently-formed title.

Maybe this is a good option for a user setting for advanced users who know what they are doing, but this could really mess up newbies.

I don't know much about the exact causes of the load problems here, but it looks kinda like that following "close" nodelinks off into oblivion (or other cool and interesting nodes) isn't the problem. Without that feature (or having that feature an extra click away), you'd break one of the neatest parts of everything2, IMO.