using CPDF before or after PSO

Question

using CPDF before or after PSO

maadjordan opened this issue 5 years ago · 16 comments

referring to pdf tool called CPDF at https://github.com/johnwhitington/cpdf-source
I've tested the following command line:
cpdf -compress -no-preserve-objstm -remove-duplicate-streams -fix-prince -remove-duplicate-fonts -remove-metadata -no-embed-font -squeeze "input" -o "output"

I got some saving occasionally by using it before PSO but using it after will rarely make any same.

maadjordan commented 5 years ago

studio ?

Answer 1 · 2019-06-19T15:16:18.000Z

Thank you for this feedback!

I'm closing thus issue because I can't see any immediate action which could be taken to improve pdfsizeopt. Feel free to reopen it if you have specific recommendations.

Answer 2 · 2019-06-19T15:17:38.000Z

BTW Multivalent can be found here
https://github.com/tsibley/multivalent-tools

Answer 3 · 2019-06-26T10:00:31.000Z

additional benefit of using CPDF it can repair and reconstruct bad or damaged streams that resulted into pdfsizeopt make a succesful optimization instead of generating error.

Answer 4 · 2019-06-26T14:35:48.000Z

@maadjordan say> cpdf ...

Very useful info!

Answer 5 · 2019-06-30T10:23:07.000Z

Also the following:
1- Qpdf has reconstruction abilities.. to be considered too ..
2- using cpdf for optimizing and compressing prior PSO will reduce resulted saving.

Answer 6 · 2019-06-30T10:42:08.000Z

@maadjordan say> Qpdf has reconstruction abilities

Details in the studio, please.

Answer 7 · 2019-06-30T11:59:37.000Z

@maadjordan say> studio ?

This is a figurative expression, an example is expected.

Answer 8 · 2019-06-30T13:06:05.000Z

i am still testing as i am getting different results

Answer 9 · 2019-06-30T15:16:51.000Z

sample file: http://www.arvindguptatoys.com/arvindgupta/paper-polyhedra.pdf
Original 2,626,880 bytes
PSO 342,473 bytes
CPDF 344,856 bytes
PSO+CPDF 340,916 bytes
CPDF+PDF 342,051 bytes

this file cpdf after pso is best

Answer 10 · 2019-06-30T15:24:08.000Z

@maadjordan say> CPDF+PDF 342,051 bytes

Maybe CPDF+QPDF? And options QPDF?

Answer 11 · 2019-06-30T15:26:48.000Z

currently using

cpdf -compress -no-preserve-objstm -remove-duplicate-streams -fix-prince -remove-duplicate-fonts -remove-metadata -no-embed-font -squeeze "%i" -o "%i.cpd"

pdfsizeopt --v=20 --use-pngout=No --use-image-optimizer=ECT --use-multivalent=YES --do-fast-bilevel-images=Yes --tmp-dir=d:\temp "%i" "%i.pso"

Answer 12 · 2019-06-30T15:36:00.000Z

@maadjordan say> currently using

Not. Interested "Qpdf has reconstruction abilities".

Answer 13 · 2019-06-30T20:23:06.000Z

Just for the record, if I use qpdf (a tool that pdfsizeopt already uses/recommends, at least for encrypted files) to uncompress the original PDF file before feeding it to pdfsizeopt, but then I can get (frequently, but not always) some extra compression gains instead of simply running pdfsizeopt directly. I suspect that the gains (sometimes, substantial, other times, smaller) come from the recompression with deflate at higher levels than the originals. I noticed this fact mainly with books from Packt and O'Reilly, but it also applies to other publishers too. I modified my own copy of pdfsizeopt to never use what Péter calls "low effort" (or something similar in the source code) and always use level 9 of deflate. If I recall correctly, Péter decides, sometimes, to use level 9, sometimes to use level 3. The "standard" is to always use level 6. My copy has also some slight tweaks to be more lenient with defective palette sizes and so on... I will add more comments as soon as I have some more free time. Regards, Rogério Brito. Em dom, 30 de jun de 2019 12:36, звездочёт <notifications@github.com> escreveu:

…

@maadjordan <https://github.com/maadjordan> say> currently using Not. Interested "Qpdf has reconstruction abilities". — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#119?email_source=notifications&email_token=AABTZMM6SDQ3NLDBO2TTPLDP5DHGDA5CNFSM4HZKTIOKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODY4OPVQ#issuecomment-507045846>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AABTZMLGQ7DPYUZD37GK42LP5DHGDANCNFSM4HZKTIOA> .

Answer 14 · 2019-07-01T03:18:03.000Z

@rbrito say> I will add more comments as soon as I have some more free time.

Good, thank.

Answer 15 · 2019-07-04T12:03:39.000Z

I modified my own copy of pdfsizeopt to never use what Péter calls "low
effort" (or something similar in the source code) and always use level 9 of
deflate.

@rbrito, please file a separate issue for this. pdfsizeopt should be able to recompress all flate streams with effort 9, and probably that should be the default.