Shotnest · Library
Compare text with repeated spaces
Red blue with two spaces is not exactly equal to Red blue, but both collapse to red blue, so normalized equality is Yes and the token overlap is 1.
Three levels of sameness
Exactly equal compares the raw strings character by character, so a second space makes Red blue different from Red blue. Equal after normalization first lowercases and folds accents, then replaces every run of spaces, tabs or line breaks with one space and trims both ends. Both inputs become red blue. Token overlap then compares the distinct words, here an identical set, giving 1.
The same collapse explains other pairs. A tab and a trailing line break, as in Red, tab, blue, newline, normalize to red blue as well. Café and cafe also become equal, because accent folding happens before whitespace is collapsed.
The letter table used by search runs here as well, so full-width letters and the ideographic space are converted before whitespace is collapsed: ＲＥＤ ＢＬＵＥ normalizes to red blue and is equal to it.
Try the original example
- First text
- Red blue
- Second text
- Red blue
- Exactly equal
- No
- Equal after normalization
- Yes
- Token overlap
- 1
Decide whether spacing matters
- Read Exactly equal first; No means at least one character differs.
- Read Equal after normalization; Yes means the difference was case, accents or whitespace.
- Ask whether the layout carries meaning, as it does in code, poetry or a table.
- If it does, compare the original screenshots rather than the extracted text.
- Red, tab, blue, newline vs red blue · normalized
- Yes
- Café vs cafe · exact
- No
- Café vs cafe · normalized
- Yes
Mistakes with collapsed spacing
- Calling two code snippets identical when only the normalized line says Yes; indentation may differ.
- Expecting the search box to collapse spaces in the same way; on this site it does not.
- Forgetting that text recognition can add or drop spaces between columns.
- Reading normalized Yes as proof that the spelling is the same; Müller and Muller are equal after folding, although they may be different names.
Equal text, possibly different pictures
Normalized equality describes the extracted words only. Two screenshots with equal text can still differ in layout, colour or a detail that text recognition missed.
In Shotnest's cleanup the text check uses letter-only words, so digits are ignored there as well; the picture fingerprint is what separates two near-identical screens.