Ebay no longer supporting wildcard in searches - Why?

Question for computer scientists:

Why would ebay drop the wildcard (asterisk *) search feature? Is there some programming science that I’m not aware of?

I collect old documents from Chicago. Like this:

chi* 18*

This gives me any of these combinations:

chi
chicago
chitown
chicag
chigaco

plus anything dated 1800 to 1899.

It is a saved search so when I get a match ebay sends me an email.

Per ebay’s suggestion I’m now supposed to search for all possibilities specifically in parentheses, like this:

chi (1800,1801,1802,1803,1804…1899)
chicago (1800,1801,1802,1803,1804…1899)
chitown (1800,1801,1802,1803,1804…1899)
chicag (1800,1801,1802,1803,1804…1899)
chigaco (1800,1801,1802,1803,1804…1899)

The result is appx 495 searches and 495 emails.

My quesion is, what is the programming science behind such a code change? A wildcard search vs a specific search still results in all the items in the database being searched.

I don’t get it.

This doesn’t answer your question, but might make your search easier:

should bring up what you’re looking for. I believe it will search for anything with “18” in it even without the asterisk.

Thanks for the reply and the suggestion. The results are not the same, I just tried it (albeit slightly different)

I searched for this:

“uss” +(1930,1931,9132,1933,1934,1935,1936,1937,1938,1939)

And got 938 results (in the stamps category).

Next I searched for: 193 (“uss”) and got 6 results.

Usually I search for “uss” +(193*,194*) for all USS Navy ship related items dated 1930’s and 1940’s. Clearly the person making decisions at ebay is not thinking clearly.

This is most likely a policy-driven change, not a technical issue. It is possible that they removed wildcard searching because it didn’t scale well. They may be using a new index that gets great performance at the expense of fuzzy matching, but they could still offer the old way to registered users or otherwise limit it. For instance, you’ll notice that when you search completed listings in a web browser, you can’t limit it to only sold items, but with the android app, you can. To me, that sounds like they want to prevent screen scraping, but realize that it’s a convenient feature.

There are two key aspects of wildcards. They cost $ to implement and maintain. Programmer time isn’t at all cheap and more complex data structures have to be implemented and maintained. The searches themselves take more computing cycles. The second is that very, very few users use them. So you’re laying out a lot of money relative to the number of people who will use it.

If ordinary folks used them, then they’d be implemented in a lot more search engines.

As a hard core computer geek, not being able to use wild cards is maddening. But my needs don’t matter. Most tech is driven by catering to the masses as cheaply as possible.

(But this is no excuse for web sites that default to “or” on keyword searches. Even the masses rarely want that.)

Regarding your two points that I bolded:

  1. The wildcard asterisk has been in place for over a decade, and it functions perfectly…
  2. Every record in the database has to be searched; a specific search string, or wildcard search (i.e. all 10,000 records have to be searched either way)

So given that do you still think that is a logical reason to eliminate the wild card? Correct me if I’m wrong. (Btw, maddening is exactly right.)

No doubt, they’ve changed their search algos in the last couple weeks.

I frequently search Ebay for parts for my backpack-carried HF radio aka a manpack.
Searching manpack now returns auctions including ‘man pack’ which has about 1200 hits at the moment, including 1000 that have no relevance what I intended. A search for “manpack” (in quotes) however drops the number to about 200 which are the desired results.
Very annoying and I knew I wasn’t going to be the only one who was irritated.

This is not correct. Without going into too much detail, database searches work by indexing searchable words separately from the records themselves. It’s a bit like looking up a term in the index of a book - far more efficient than reading the entire book every time you want to find something.

There are a lot of ways to do simple suffix wildcard searches. Gets too complicated to explain many of them and I have no clue as to how eBay did it. (And if you’re doing infix or prefix wildcards, the data structures are at entirely higher level of complexity.)

One of the simple ones to understand are tries. Start with the Wikipedia article and note the comparisons to other data structures and go as deep into as you want.

Without wildcards, you can do a lot more hash based stuff which is usually faster and takes up a lot less storage space.

As to the issue of “has been in place for over a decade”: The overwhelming cost of a piece of software is maintaining it. It’s never like write it once and forget about it. Day after day, you have to keep it going. New bugs are found. New issues arise that have to be dealt with. E.g., you switch over from disks to solid state storage and all the timing assumptions change. Even keeping up with growth means that a solution that worked well at one scale is a bad idea at a larger scale. Throw in things like the people who wrote the code might be long gone and no one around knows the minutiae of the internals and the costs get very large very quickly.

As long as the mass of users don’t care about the poor quality of search results, expect continued declines. (As we are seeing with Google’s downhill slide.)

Sounds impressive but I can’t imagine that eliminating the wildcard - and thus as you suggest - the cost of its upkeep, is a recommended formula for reducing overhead. Take a read around the interwebs from all the users who are flabbergasted that ebay would eliminate such a valuable search tool. The mission of ebay is: Sell. Ergo… wildcard = more items found = more sales.

Wise installations will tune their machines, and plan upgrades to meet growing demands. Not turn off features.