4 ms·

Based on the first example, it seems a wee bit more complicated than that: fuzzysearch('twl', 'cartwheel') // <- true The string 'cartwheel' does not cont

by sync 12y ago

Based on the first example, it seems a wee bit more complicated than that: fuzzysearch('twl', 'cartwheel') // <- true The string 'cartwheel' does not contain the string 'twl' directly.

bshimmin 12y ago
I think new RegExp("twl".split("").join("(.+)?")).test("cartwheel") does approximately the same, no?
- Guillaume86 12y ago
  Yes it's the same according to the source. You can also use ".*" to replace "(.+)?" in your regex.
  - mistercow 12y ago
    It's only the same as long as your search string doesn't contain any special characters as far as regex is concerned.
    - Guillaume86 12y ago
      Yes good catch. It's going to be a lot slower anyway. Still does a good job at explaining the matching algorithm concisely.
- solox3 12y ago
  I guess fuzzy search also allows for substitutions like "csrtwheel", but neither you nor the OP does this.
  - kelseyfrancis 12y ago
    Yeah, anyone thinking of using this should probably instead just look for an existing implementation of something like the Damerau–Levenshtein distance.
    - logicallee 12y ago
      >anyone thinking of using this should probably instead just look for an existing implementation of something like the Damerau–Levenshtein distance. "anyone" with over 400 bytes, you mean.
    - jgalt212 12y ago
      I don't know how computationally intensive OP's algo is, but I've found that in real world use cases Damerau–Levenshtein distance algo has slowed down a number of our jobs (and we aim to avoid it is possible). Our jobs are real-time where the user is awaiting a response. So probably not limiting factor for jobs that can be run on your own time.
    - bevacqua 12y ago
      OP here. My code is meant to be used for client-side bursty usage (like filtering results for an autocomplete list), which demands the code to be quick and small. In this use case, it's not necessary to figure out how similar strings are, because it's completely irrelevant. If you're building your own search engine, which you probably shouldn't, I'd look elsewhere.
      - magoon 12y ago
        > If you're building your own search engine, which you probably shouldn't, I'd look elsewhere. Advice not taken.
      - kelseyfrancis 12y ago
        > In this use case, it's not necessary to figure out how similar strings are, because it's completely irrelevant. I don't follow, since that's pretty much the definition of "fuzzy search". Your algorithm considers the query to be similar to the searched text if the latter contains all the characters in query in the same order. If the use case is just to allow users to type the first few letters of each word or something, that algorithm works great, so advising "anyone" to look elsewhere didn't make sense. If users expect to be able to make typos other than leaving out characters, though, it doesn't work, and that's often what people think of when they see "fuzzy search". I just wanted to point out that you could use Levenshtein + a BK-Tree or something and get those results if desired, though, yeah, it'd definitely be more complicated, but I doubt it'd be too slow.
- bevacqua 12y ago
  Except egrigiously slower https://github.com/bevacqua/fuzzysearch/pull/6#issuecomment-77263900 https://github.com/bevacqua/fuzzysearch/pull/6#issuecomment-...

by sync 12y ago

Based on the first example, it seems a wee bit more complicated than that: fuzzysearch('twl', 'cartwheel') // <- true The string 'cartwheel' does not contain the string 'twl' directly.

bshimmin 12y ago

I think new RegExp("twl".split("").join("(.+)?")).test("cartwheel") does approximately the same, no?

Guillaume86 12y ago
Yes it's the same according to the source. You can also use ".*" to replace "(.+)?" in your regex.
- mistercow 12y ago
  It's only the same as long as your search string doesn't contain any special characters as far as regex is concerned.
  - Guillaume86 12y ago
    Yes good catch. It's going to be a lot slower anyway. Still does a good job at explaining the matching algorithm concisely.
solox3 12y ago
I guess fuzzy search also allows for substitutions like "csrtwheel", but neither you nor the OP does this.
- kelseyfrancis 12y ago
  Yeah, anyone thinking of using this should probably instead just look for an existing implementation of something like the Damerau–Levenshtein distance.
  - logicallee 12y ago
    >anyone thinking of using this should probably instead just look for an existing implementation of something like the Damerau–Levenshtein distance. "anyone" with over 400 bytes, you mean.
  - jgalt212 12y ago
    I don't know how computationally intensive OP's algo is, but I've found that in real world use cases Damerau–Levenshtein distance algo has slowed down a number of our jobs (and we aim to avoid it is possible). Our jobs are real-time where the user is awaiting a response. So probably not limiting factor for jobs that can be run on your own time.
  - bevacqua 12y ago
    OP here. My code is meant to be used for client-side bursty usage (like filtering results for an autocomplete list), which demands the code to be quick and small. In this use case, it's not necessary to figure out how similar strings are, because it's completely irrelevant. If you're building your own search engine, which you probably shouldn't, I'd look elsewhere.
    - magoon 12y ago
      > If you're building your own search engine, which you probably shouldn't, I'd look elsewhere. Advice not taken.
    - kelseyfrancis 12y ago
      > In this use case, it's not necessary to figure out how similar strings are, because it's completely irrelevant. I don't follow, since that's pretty much the definition of "fuzzy search". Your algorithm considers the query to be similar to the searched text if the latter contains all the characters in query in the same order. If the use case is just to allow users to type the first few letters of each word or something, that algorithm works great, so advising "anyone" to look elsewhere didn't make sense. If users expect to be able to make typos other than leaving out characters, though, it doesn't work, and that's often what people think of when they see "fuzzy search". I just wanted to point out that you could use Levenshtein + a BK-Tree or something and get those results if desired, though, yeah, it'd definitely be more complicated, but I doubt it'd be too slow.
bevacqua 12y ago
Except egrigiously slower https://github.com/bevacqua/fuzzysearch/pull/6#issuecomment-77263900 https://github.com/bevacqua/fuzzysearch/pull/6#issuecomment-...