Hacker News Reader

No, you're wrong. It's not with the data. It's with Apple's ability to search. Like I said below, look up "Pier 39 San Francisco". This is a dead simple query, extremely specific, and they point you to a really bad area of SF. This isn't Tom Tom data, it's how Apple resolves this query to something useful.

Search is Google's domain expertise. Apple needs to basically start from scratch and figure out how to create a search engine that will return relevant data based on queries.

It's the same thing with Apple's App store search, it's also terrible. But because we have no choice, we are forced to use it, and hopefully it will improve. But because Google's is so nearly perfect and getting better all the time, why would someone go ahead and use Apple's terrible search engine to help them improve their search?

Apple Maps also doesn't handle small spelling mistakes (transposed letters) well. That's been one of the most annoying things.

Having played with a toy geolocation feature detector for text, I can safely say it's extremely hard to do well.

My first attempt is inspired by Peter Norvig's spelling corrector using a corpus built from the places2k.txt file, but it gets confused when the words are transposed. My guess is that Google is using some sort of n-gram based approach for maps and search, where as their little sidebar map is similar to what I built. (Or only applied to the most clear-cut n-gram—a little while ago there was a search for the movie "The end of the world" that would return a definitive date, which went slightly viral.)

Try searching in Google for "San Francisco" versus "Francisco San." Maps is fine, the search is fine, but only the first version brings up the little related info sidebar on the search page.

I'm surprised that Apple had such a hard time dealing with search, since they have some experience with Siri. I should have trusted this guy's opinion more:

http://wesnerm.blogs.com/net_undocumented/2011/10/siri-techn...

> Try searching in Google for "San Francisco" versus "Francisco San."

On the other hand, that's not a mistake you can make, because "San Francisco" in the minds of people really is "Sanfrancisco", a single word. Why would anybody spell "San Francisco" as "Francisco San"? In fact Google should differentiate heavily between these 2 terms, as they aren't similar at all.

Google also does a lot of other neat things, like translations for city names. Searching on google.com for "Bucuresti" (my town) brings up the sidebar for "Bucharest".

And on GMaps, I'm never afraid of using "str", "strada", "soseaua", "sos" which are different Romanian words and abbreviations for "street". The results aren't the same in these cases (nor they should be actually) but because Google also searches within the addresses of businesses it's usually able to give me an accurate answer to what I mean.

That's true about San Francisco versus Francisco San but I wonder how well Apple Maps does with "4485 Belmont Portland" versus "Portland 4485 Belmont" and the like.

It gets more complicated when you realize "oh, yeah I need to add 'NE' because it's in the Northeast part of the city". So now it's "Portland 4485 Belmont NE". If you were to think about it before you started typing, it'd be "4485 NE Belmont Portland" but correcting on a phone is so painful that you want to just keep typing.

I'm using a Galaxy Nexus and have been very happy with the standard Maps app. Last weekend the family was walking on trails and it got us back to the trailhead using walking directions. Neat. I've been less than thrilled with how poorly the phone seems to do with finding and keeping a GPS signal, particularly in a car on the freeway in areas I don't know.

Ah, Portland is lucky. Here in Seattle, we have roads that are 1235 NE 40th St (for E-W streets) as well as 12345 University Way NE (for N-S avenues). The latter, as you would suspect, totally breaks Apple Maps.

That is on top of the issue of searching for something like "1st and University", which takes me to some L'ecole in Paris, France rather than the intersection that is a block from me. It's as if they don't factor in your current location at all. I know on the desktop I frequently look at maps of faraway places, but I'd suspect that people tend to look for things near them while on mobile devices.

Apple Maps doesn't know what you mean when everything is spelled perfectly, they have a lot to do before worrying about typos.

Surely you agree that your example is a non-standard address. The things are more important to get right: postal addresses, searches for company names, airports, schools, searches for intersections. Of course you want to get the other things right like, parks (what's the pinpoint location of a park?) and things like pier 39. But those errors don't make Apple Maps a non-starter. I've been using it in the Bay Area with 99% success.

This isn't a pass for Apple Maps- the last 1% is the most difficult and it's the difference between so-so and great. Apple has in the past been known to make things with the ultimate attention to detail and things that "just work" so if anything you could argue that they shouldn't have launched something that was below this bar, but apparently there were other business considerations.

BTW: From the Pier 39 website, "If you are using a GPS Navigation System, input: 2 Beach Street, San Francisco, CA 94133" http://www.pier39.com/Information/index.htm

If you need the user to visit the website of their intended destination to figure out how to search for it then you need to hit the drawing board.

Well, that's the point I'm trying to make. It's not a hit the drawing board problem, it's a refine the algorithm problem. How many "if you google this string, you can't find the right site" problems have there been in the life of Google search? They continue to refine pageranking don't they?

'Apple has in the past been known to make things with the ultimate attention to detail and things that "just work"'

I don't know why people continue to believe this. Apple has released versions of the Finder with amateurish bugs that delete files[1] and versions of Mail that randomly delete messages[2]. Several versions of Mail on iPhone send hundreds of copies of a message when emailing a link from Safari[3].

The Maps fiasco is not a departure from software excellence on the part of Apple, but just more of what we should expect by now: shiny stuff that doesn't work.

[1]http://tomkarpik.com/articles/massive-data-loss-bug-in-leopa... [2]http://discussions.apple.com/thread.jspa?messageID=12758081&... [3]http://lee-phillips.org/iphoneUpgradeWarning-4-2-1/

"shiny stuff that doesn't work"

Give me a break. Name an international company with 100% success in user happiness and usability with their product(s). Go ahead, try it.

Also, don't act like Android hasn't had faults and failings on their own. Or Windows.

Things aren't ever perfect. The only time there is a real binary system is CS. The world is gray, and Apple has been consistently better than bad. They've been good for me for the most part, with some minor annoyances sure, but to say that their software is "shiny stuff that doesn't work" just shows that you aren't really well versed in the domain that the hyperbole is supposed to explain.

I didn't claim that others' products were perfect. That would be strange. My point was that Apple enjoys an undeserved reputation for quality.

"Apple has been consistently better than bad"

I think my links show that this is just not true, but that might depend on what you mean by "consistently".

Their track record shows a cynical, almost sinister pattern of choices on their part: hoard $120 billion in cash while shipping software with a beautiful visual design that, under the glossy surface, is unreliable, inadequately tested, and, in some cases, demonstrably hobbled together by inexperienced or incompetent engineers: shiny stuff that doesn't work.

This is a deliberate choice, since with all that cash they could hire the best talent and make it policy to ensure that nothing gets out the door that would earn a CS undergraduate a failing grade. Instead, they've chosen to invest in maintaining the illusion that their software "just works" and is empowering, while harboring an increasingly obvious contempt for their customers.

ref 1 dates back to 2007, ref 2 is an isolated 'bug report' which no one else seems to corroborate and ref 3 seems to be written by you. You're not making any kind of case here for Apple being overrated.

"ref 1 dates back to 2007"

And?

"ref 2 is an isolated 'bug report' which no one else seems to corroborate"

It's one link to a report of a well-known, real problem. Apple Mail botched IMAP semantics just as the Finder botched file copying semantics. Both problems are reliably absent from real, mature software. And even in the one link I provide, several users are reporting disappearing messages. I would call that "corroborated", wouldn't you?

"ref 3 seems to be written by you"

It was, so you can be sure it's true.

Even with company names, they do a pretty poor job. Though it is getting better.

For example, a month or so back a search for "Northern Brewer" (homebrew store chain) made when I'm physically only about 2 miles away from their Milwaukee location would return only the location of their flagship store in the Twin Cities. Which I suppose is technically a correct result, but it's a little galling to have regressed to a mapping app that doesn't even understand concepts as basic as, "When I just enter a business's name all by itself, it just might be reasonable to assume I was looking for a local place."

Some are still really bad. A search for the Miller Brewery's bar, Miller Inn, produces a single result for a business called Fabric Inn in Miller, South Dakota.

Is there any particular reason for the dickish tone? Are we in grade school?

funnily enough when typed pier 39, the first auto suggestion was "pier 39" ( the right one) - the second one was "pier 39 san francisco" which i what i thought was the correct one and tapped on , but turned out to a restaurant named 39 pier.

"why would someone go ahead and use Apple's terrible search engine to help them improve their search?"

Because at the end of the day I want to make choices that result in more flexibility down the line. I do not want to live in a world where Apple is the only OS provider for mobile devices. Likewise, I don't want to live in a world in which Google is the only reasonable choice for map searches.

So is your strategy to always use the worse of two services to ensure competition for the future?

This. Arguing to use an inferior product so that better products don't succeed "too much" isn't how things work.

Markets work this stuff out, you don't necessarily have to balance market share and consequences in your head (not that you as an individual really matter that much in the end anyway!). Just enjoy the products you like.

Nope. Apple Maps isn't particularly worse for me. The directions have been fine, and I like turn-by-turn. I miss transit directions quite a bit. I thought about what I was missing and I thought about how dominant Google is in search, and I decided the tradeoff was worth it for me.

I wouldn't make fun of someone for making the other choice either.

That Pier 39 San Francisco search is an Australia-desert-type disaster waiting to happen to tourists.

Speaking of tourist disasters in SF; when I was there back in May, there were a few blocks somewhere around (IIRC) Post/Van Ness where the GPS in my Galaxy Nexus would actually pick up a mile or so off from where I actually was. Realized what was happening right away and still know how to read maps[0]; but could easily be a problem for anyone who isn't paying attention to everything.

[0] Semi-related: whoever decided to stamp street names into SF sidewalks: thank you.

> No, you're wrong. It's not with the data.

It seems to be both. I tried two locations in the Apple store, and the data for both was incorrect:

* One gave a location about 10-15 miles away (with a pop-up showing details of the correct venue). I checked the correct location, and nothing was marked on the map. (Image: http://i.imgur.com/FHVax.jpg - the proper location is just above the 'd' in Cambridge)

* One was found correctly, but the listed information (phone, website etc.) was mostly incorrect.

* Satellite images are of a much lower resolution than Google Maps in some areas (e.g. my house!)

Google Maps is vector based. Are you perhaps talking about the original "Maps" app on iOS 5 (and below) which relied on Google data?

On Android, you can select regions of the map to save for offline use and the file sizes for entire cities are only a few MB.

I don't believe a "save for offline use" option is available for Google Maps for iOS yet.

> If apple is able to partner with a company with better data, these types of problems should not be nearly as bad.

There is none. As far as I know, there are two major providers for map data. One is TeleAtlas, which is now owned by TomTom. The other one is Navtec from the US. TeleAtlas is considered slightly better in terms of quality, as far as I know.

Google Maps used to use a combination of both providers. However, now they also have their own data from street view cars in many places.

Navteq is owned by Nokia, btw.

And the TomTom data is fine. When Apple maps fails to find a POI, I open my TomTom app, and it always knows it - offline too. Apple's screw up is giving TomTom a bad name...

If you want to try out some other vector based map system there's Open StreetMap. I have no idea about how functional the iPhone apps are but OsmAnd for Android is able to cache all map data for Sweden in 200MB.

It is then able to do routing on that data for both walking and going by car, both from search and random positions pointed at on the map. Not to mention that it has information about resturants so that you can plan where to stop for lunch when going on longer trips, all of which without a data connection.

Google Maps are vector based as well...

Not on my cell phone.

I think you're mistaken. They are on Android so it seems odd they wouldn't be.

The ability to zoom to arbitrary levels with no pixellation seems to imply it most definitely is vector.

On iOS it appears they're a mixture of bitmap tiles and vector text. I definitely see pixelation of roads at certain zoom levels.

Can you give steps to reproduce? I was looking specifically at road outlines and they seemed like they were definitely vector.

Satellite views are obviously bitmap.

I don't have an iOS device, but here are my steps to reproduce on Android. Open a map. Turn on airplane mode. Zoom in. See pixels

Here's a screenshot I took after I enabled airplane mode: http://i.imgur.com/wc5Ie.png

The pixellation here is your phone waiting to download the higher-res vector tile. While it waits for the higher-res tile it draws a bitmap representation of the lower-res tile "above" it.

Surely if it's vector done right they can draw the lower detail vector tile at the higher bitmap res?

No because the amount of detail will change at each zoom level. Smaller roads and a higher number of local details appear.

They're not infinitely vectors. They're still tiles. If they were to use the zoomed in tile, when zoomed out, even with vectors, you'd get something unrecognizable.

It's the same reason that fonts and icons are brought up when people espouse the "vectors will save everything"! They do a lot of nice things but for something like Google Maps, I can't imagine how they'd implement "true vectors" in a practical phone-applicable way.

No, they're vectors, intelligent vectors that add and remove features at different zoom levels.

http://googleblog.blogspot.com/2012/12/google-maps-is-now-av...

I am using an Android device. And I do clearly see pixelation when I zoom in, before the new tiles load.

Here's a screenshot I took after I enabled airplane mode: http://i.imgur.com/wc5Ie.png

Which version of the Maps app are you using?

I don't see these blurred pixels on my phone, but note that this kind of rendering can also happen with a vector based map: The application has to convert the vector information to a bitmap anyway, for a given zoom level. When zooming in, it is possible that the app shows a zoomed-in version of this bitmap while it is generating a new one in the background.

By the way, it is not necessary to answer every single comment with a link to your screenshot. It is annoying to see the same thing repeated 5 times when reading the thread.

I'm using 6.14.1.

I answered all the relevant replies individually because I thought that would make those people more likely to see it. Sorry for annoying you.

Well now you have five answers like mine :)

Same version here though the UI looks very different. I can actually reproduce what you describe but only at low zoom levels, which makes sense: the temporary display of a bitmap at the wrong zoom level is more apparent when the phone needs more time to render a lot of vector data (there is less data to render when zoomed-in).

Curiously, I wonder what they do then because they definitely have the vector data in the backend. Google Maps for Android has been vector based for 1.5 to 2 years now.

I'm using google maps for android. When I zoom in, I can clearly see the pixelation of the tiles before the new tiles load. The google play store doesn't show any updates for the app. Where are you getting your information?

Here's a screenshot I took after I enabled airplane mode: http://i.imgur.com/wc5Ie.png

I have this on an old HTC phone. The UI is totally different, I think they have different layout controls for the "newness" of the phone. Having said that, it's possible the vector data is still being used, but renders to bitmap when zooming for some reason. I can't recall this happening on my S2 though (dead now, can't test)

But we're talking about the new Google Maps app for iOS, which is not the same as Google Maps for Android (its a newer ground-up design with different features.)

Perhaps you need to install the App we're all talking about then. It's most definitely vector based.

I thought we were talking about the Google Maps app. Because that's what I'm talking about.

Here's a screenshot I took after I enabled airplane mode: http://i.imgur.com/wc5Ie.png

Well, for starters, you're not using an iPhone and that's not what the Google Maps app for iOS looks like.

Regardless, your screenshot shows no evidence that the maps aren't vector based. You simply don't have the high-resolution data for that area and Google's UI is showing that to you by burring it. Apple Maps doesn't blur, it just doesn't show you any details. I like knowing that there is information I am missing, the blur is handy.

Which means you must be using it in a browser, on a Java phone/blackberry, or in iOS 5. Try the new iOS app or an Android phone.

I'm using the maps app on and Android phone. When I zoom in, I see the pixelation before the new tileset loads.

Here's a screenshot I took after I enabled airplane mode: http://i.imgur.com/wc5Ie.png

It's definitely vector-based. What you're seeing is that the vector tile gets rendered to a bitmap for display. If you zoom in quickly, rather than re-render the bitmap over and over, it just scales the pre-rendered bitmap. When the new zoomed-in tile loads, then it re-renders that tile.

Also, even though it's vector-based, there's still different tiles for different zoom levels. For example, with different levels of detail.

The new app is vector based and has turn by turn directions.

Google refused Apple access to both of those features in the original "Maps" app (in iOS <= 5), but they conveniently included them in their new Google Maps app. :)

Google refused to allow Apple access because Apple was unwilling to allow more Google branding and Latitude support. It was a breakdown in negotiations, not either side being selfish.

This comment is clearly unbiased (and correct), which makes it all the more annoying that your comment (elsewhere in this thread) about Google's early troubles with search in their own app store has been being voted down.

> Google Maps are vector based as well...

src: http://googleblog.blogspot.com/2012/12/google-maps-is-now-av...

Google Maps have been vector based for a while.

At least, on Android.

The new Google Maps app for iOS is also vector-based.

> I've used google navigation since the first day the alpha was leaked. It has only been in the last year that nav has gotten so good (used to have maddening ui problems with the zoom going crazy)

Really? I thought maps was better in iOS 3 than iOS 5. Maps in 3 was optimized to be quick even over EDGE and never let go of a set of directions mid trip if network was flaky. iOS 3 maps was actually faster and more reliable. (Still had my original iPhone)

I'ved used Android's "Navigation" app on different older phones without a data plan (only WiFi) and they've been surprisingly good about setting up a route and letting me follow it.

(I can't change my route mid-trip, which was maybe what you were saying.)

as others have pointed out gmaps is vector based.

Also gmaps (for android at least) allows you to download map data for offline/online use. So to say theres a huge dependence on data connection is not true.

For me the main issue is StreetView integration. Apple wont have that any time soon.