9

The word bið is not the same as bid because one is Old English for be and the other is an etymologically unrelated word that you can find in any dictionary.

The what the eth here is that search treats ð and Ð the same as d and D. Just look at this search:

bid is in the search results for bið

Eth is used in many Old English words and is certainly not the same as d. Another reason this needs to be fixed is because eth is used in IPA notation to represent the voiced dental fricative (and is not the same as d there either).

While this may affect other sites, I have chosen to report it here (instead of MSE) because it affects us more than the vast majority of sites on the network.


Right now, one workaround is to use the Data Explorer:

SELECT id AS [Post Link] FROM posts WHERE body LIKE '%bið%';

And another is using Google:

"bið" site:english.stackexchange.com

But I would like to see this fixed in our search because both methods have drawbacks.

3

Yours is an interesting point, but I’m not wholly sure we want this fixed.

The technical explanation for what I believe is happening is probably going to either confuse or bore the pants out of most people reading this, but here goes anyway.

I’m pretty sure that Stack Exchange searches are done using something called the Unicode Collation Algorithm at level-one equivalence, also called the primary level or strength.

Primary strength here corresponds to “alphabetical” ordering only. That means that case and diacritics don’t count, just the "basic" letters themselves, and indeed many code points wind up being equivalent there.

In particular, at the UCA-1 level, all these count as the same letter:

U+0044 ‭ D  LATIN CAPITAL LETTER D
U+0064 ‭ d  LATIN SMALL LETTER D
U+00D0 ‭ Ð  LATIN CAPITAL LETTER ETH
U+00F0 ‭ ð  LATIN SMALL LETTER ETH
U+010E ‭ Ď  LATIN CAPITAL LETTER D WITH CARON
U+010F ‭ ď  LATIN SMALL LETTER D WITH CARON
U+0110 ‭ Đ  LATIN CAPITAL LETTER D WITH STROKE
U+0111 ‭ đ  LATIN SMALL LETTER D WITH STROKE
U+0369 ‭ ○ͩ    COMBINING LATIN SMALL LETTER D
U+1D30 ‭ ᴰ  MODIFIER LETTER CAPITAL D
U+1D48 ‭ ᵈ  MODIFIER LETTER SMALL D
U+1D9E ‭ ᶞ  MODIFIER LETTER SMALL ETH
U+1DD8 ‭ ○ᷘ  COMBINING LATIN SMALL LETTER INSULAR D
U+1DD9 ‭ ○ᷙ  COMBINING LATIN SMALL LETTER ETH
U+1E0A ‭ Ḋ  LATIN CAPITAL LETTER D WITH DOT ABOVE
U+1E0B ‭ ḋ  LATIN SMALL LETTER D WITH DOT ABOVE
U+1E0C ‭ Ḍ  LATIN CAPITAL LETTER D WITH DOT BELOW
U+1E0D ‭ ḍ  LATIN SMALL LETTER D WITH DOT BELOW
U+1E0E ‭ Ḏ  LATIN CAPITAL LETTER D WITH LINE BELOW
U+1E0F ‭ ḏ  LATIN SMALL LETTER D WITH LINE BELOW
U+1E10 ‭ Ḑ  LATIN CAPITAL LETTER D WITH CEDILLA
U+1E11 ‭ ḑ  LATIN SMALL LETTER D WITH CEDILLA
U+1E12 ‭ Ḓ  LATIN CAPITAL LETTER D WITH CIRCUMFLEX BELOW
U+1E13 ‭ ḓ  LATIN SMALL LETTER D WITH CIRCUMFLEX BELOW
U+2145 ‭ ⅅ  DOUBLE-STRUCK ITALIC CAPITAL D
U+2146 ‭ ⅆ  DOUBLE-STRUCK ITALIC SMALL D
U+216E ‭ Ⅾ  ROMAN NUMERAL FIVE HUNDRED
U+217E ‭ ⅾ  SMALL ROMAN NUMERAL FIVE HUNDRED
U+24B9 ‭ Ⓓ CIRCLED LATIN CAPITAL LETTER D
U+24D3 ‭ ⓓ CIRCLED LATIN SMALL LETTER D
U+A779 ‭ Ꝺ  LATIN CAPITAL LETTER INSULAR D
U+A77A ‭ ꝺ  LATIN SMALL LETTER INSULAR D
U+FF24 ‭ D FULLWIDTH LATIN CAPITAL LETTER D
U+FF44 ‭ d FULLWIDTH LATIN SMALL LETTER D

There may be a way to search exact code points using some sort of quoting trick, but unless such exists, I doubt you’d be able to get SE to change how searching works just for us. I believe they use something called Elasticsearch, tuned for English only.

  • Yeah, I knew it was collation. And I think they still use Elasticsearch. It may be possible to customize the collation, but I am not very familiar with the tech to know how that would work. There might be some drawbacks to if e.g. ö and o are not the same letter, but I can't see any drawbacks if only ð and d were treated as separate letters apart from developer efforts to implement (if it's possible). – Laurel Oct 29 '17 at 23:28
  • @Laurel: the only case I know of where the current behavior might be more useful than your proposed behavior is when dealing with Anglicized spellings of Norse names. For example, typically people write "Odin" as the name of the Norse god written "Óðinn" in modern Icelandic. The Norse godess "Iðunn" may be anglicized as "Idunn" or "Idun"; although apparently even "Ithun" has been used, which wouldn't get treated the same by any normal search engine. – sumelic Oct 30 '17 at 1:11

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .