search

tavinus / pdfScale

Bash Script to Scale and Resize PDFs using Ghostscript
MIT License
242 stars 36 forks source link

Centering objects in resize mode seems to be broken #28

Closed bkroggel closed 1 year ago

bkroggel commented 2 years ago

I am pretty confident that this is not immediately related to pdfScale but rather to a change in one of its dependencies that leads to the outcome.

Expectation:

When a pdf (which in our sample case is a svg file that has been converted to a pdf) gets resized to a new file that has different dimensions than the original one the larger side of the content should fill 100% of the new page while the smaller side should be centered.

Current behaviour:

Resizeing also resizes the content as expected but they do not get centered/aligned correctly.

How to replicate

  1. Get a random pdf file (e.g. the Wikipedia Logo (svg > pdf)) Wikipedia_wordmark.pdf
  2. Try to resize the object — e.g. to a object with 700x400pt: pdfscale -r "custom pt 700 400" Wikipedia_wordmark.pdf
  3. Output Wikipedia_wordmark.CUSTOM.pdf
pdfscale v2.5.3 - Verbose Execution
Single Task: Resize PDF Paper
Dry-Run: FALSE
Input File: Wikipedia_wordmark.pdf
Output File: Wikipedia_wordmark.CUSTOM.pdf
Get Page Size: Adaptive Enabled
Method: Grep
Failed
Method: Mac Quartz mdls
Source Width: 375 postscript-points
Source Height: 64 postscript-points
Print Mode: Print ( auto/empty )
Scale Factor: Disabled (resize only)
Fit To Page: Enabled (default)
Auto Rotate: PageByPage
Flip Detect: No change needed
Run Resizing: CUSTOM ( 700 x 400 ) pts
Final Status: File created successfully

OS: macOS Monterey 12.2.1 (21D62) GS: 9.56.1 Imagemagick: 7.1.0-36 Q16-HDRI arm 20076 BC: bc 1.06

Maybe as a small addon — I have another Mac running Big Sure 11.5.1 with the identical Imagemagick/BC setup but a slightly outdated GS Version (9.54.0) where everything outputs like expected — in case that is helpful.

bastian-pnptc commented 2 years ago

Digging to the Ghostscript release notes indeed showed that a major changes has been implemented recently. It seems like the rewritten PDF interpreter has caused the issues mentioned above. https://www.ghostscript.com/doc/current/History9.htm

As a workaround adding the flag -dNEWPDF=false will force Ghostscript to still make use of the old PDF interpreter written in PostScript rather than C.

That obviously isn‘t a perfect solution as the reimplemented interpreter should do the trick as well — however it at least keeps everything as it should be for now.

tavinus commented 2 years ago

Sorry for the late reply. This is interesting.

I wonder:

  1. If there is a way to enable the new parser and get the same results
  2. If using -dNEWPDF=false on older GS versions will break the script (we need to test the version before using it if that is the case)

I will need to investigate...