Search box not working

[alberto-ardito]

Hi, I have a problem: the search box does not work for some pdfs (usually google books) even though the text is searchable and selectable.

[GitHubRulesOK]

Text can be 4 or 5 primary types

Conventional text encoded as single characters (causes a few oddities due to the way white spaces are stored as missing or need skipping)
Annotation such as form fields that are considered as comments outside the page but reference an area of a page. (even some "flatter" text)
Vector pathed texts (may be some from clear type/dvju/ms printed/SVG/emoji/alt text sources.
Pixels (images including SVG Emoji etc.)
Artwork (includes SVG and other similar letter forms)

Any of those may even have a substitute in the page that is garbled or designed for specialist decoding like Alt Text.

So without a sample the odds above are more likely to be non-searchable than truly searchable.

[alberto-ardito]

It's a matter of spaces as you say: if I search for "good" the word is not found, if instead I search for "g o o d" the word is found. I'll make do, thank you

[GitHubRulesOK]

That is most commonly 2 sources one is OCR and since you say "Google Books" that is highly likely.
SumatraPDF works on the basis the text is literally decoded as (Hello World) or very similar However OCR is single characters since that is what the C means "Character" in the singular. Then it strings them into groups (H)-#(e)-#(ll)-#(o)-##(W)+#(o)-#(r)-#(l)-#(d) may be a common spaced out scanned line.

If those gaps shown as # are a big enough - negative (spaced) value as seen between (o)-##(W) it counts as a space character so depends what scaling value Google applied.

The space can thus easily be between letters but if + positive (closer) the 2 characters will generally be considered adjoining / overlapping hence Wo become tucked tightly! The term is "Kerning" and values usually low like 1 or 2 but can easily be 100s! and then the gaps appear.

There are wrinkles such as fi may count as a single character called a "Ligature" such as I showed (ll) might be seen as one

[alberto-ardito]

An example: https://books.google.it/books?id=szSPQGdteEUC
Since in other PDF readers (such as Chrome) you can search for words without inserting spaces between letters, I suppose that SumatraPDF is interpreting certain OCR values as spaces.
Do you think instead that the OCR data is "wrong"?

[GitHubRulesOK]

Whilst "online" some PDF viewers are simply showing a reproduction of a PDF so you can edit your copy as images and in some cases they will extract text and convert to HTML I dont see where you get the PDF for SumatraPDF as that link is to googles viewer of images

Sorry I should have looked closer
so on download we can see the internals and SumatraPDF can find whole words as written by the OCR

A more conventional page such as 10

Here I remove the image from the background and we can see that as with all electronic files the text is just a binary character at a page position. PDF uses a numeric system which on occasion may be converted to Abc but in many cases does not need to be as the reader can be guided by the documents list of convert number to character for Audio and search etc.

What do I mean by Numbers well here is that start of Page 10 and luckily since it uses English for Italian (Latin-American system :-)

I know 45 is E thus Erodoto is [<0045><0072><006F><0064><006F><0074><006F>] so we can search for that number group.

[alberto-ardito]

So, is this only my problem?
The contents of the pdf in pdf-xchange are the same as your screenshot but I can't search them with sumatrapdf:

[GitHubRulesOK]

ODD
There are 3 possibilities

• Document I presume we have similar copies

• Language I tried that and not need space in my copy
  
  

• The remaining one is the version / build (I am using a current Pre-release 3.6) and that does seem to be an issue from previous 3.5.2
  

[alberto-ardito]

I've tried with the pre-release version: same result. Such a mistery

[GitHubRulesOK]

Sorry it looks like I have a newer version !! ( I usually compile my own)

I downloaded (I can only test 32bit) 16779
https://www.sumatrapdfreader.org/dl/prerel/16779/SumatraPDF-prerel-32.exe
and from dropdown on right of https://books.google.it/books?id=szSPQGdteEUC&printsec=frontcover&hl=it&source=gbs_ge_summary_r&cad=0#v=onepage&q&f=false

Scarica PDF and saved as before

But just like You I see

[GitHubRulesOK]

@kjk any news on Certification / newer pre-release