| | <<<Back 1 day (to 2018/11/09) | 20181110 |
| CanadianHusky | Hello, I was wondering if anyone had time to look at bug # 700052. It makes GS 9.25 completely unusable for quite a few color to grayscale conversions using pdfwrite device when the content has existing gradients/grayscale objects. GS 9.20 does a fairly good (not perfect but acceptable) conversion on the provided sample file. Thank you kindly | 09:07.30 |
| kens | The old versions do not reatin teh shading dictioanry on colour conversion, but render it to an image | 09:07.57 |
| | If an image is acceptable to you, then use the pdfimage8 device | 09:08.15 |
| CanadianHusky | content is required to stay as vector | 09:08.40 |
| kens | Then you'll have to wait, I don't regard this as an urgent issue | 09:09.00 |
| CanadianHusky | gs 9.25 looses a lot of objects and is a serious regression when compared with 9.20 | 09:09.10 |
| | understood | 09:09.16 |
| kens | Note that, as I said, the older versions don't reatin the shading, so it wasn't retained as 'vector' previously | 09:10.07 |
| CanadianHusky | the page content has a lot of other objects that stays fine as vector with 9.20. Of course you are correct, the grayscale part in question gets rendered to some sort of image with 9.20, with a bit quality loss, but more or less acceptable | 09:11.29 |
| | Its not elegant but because of these issues, I have added code to use 2 versions of GS at the same time depending on the task at hand, GS 9.20 for grayscale conversions and 9.23 for pdfimage operations. 9.24 and 9.25 have around 800% performance penalty on pdfwrite device on some files and I am hoping the recent commit you did about the glyph issue performs well. I could not test it yet, waiting for the official build. Th | 09:14.14 |
| | this code snippet did another approach to the grayscale conversion and on some files (also the file in the sample) works really well and preserves the content as much as possible. His logic does fail on other files where GS does an excellent job. Maybe it can act as an inspiration when the time is right to review that part of the code. http://habjan.blogspot.com/2013/09/proof-of-concept-converting-pdf-files.html | 09:23.57 |
| kens | No. We need to convert an M-in N-Out function into an M-in 1-out function | 09:25.10 |
| | Which means executing the function and sampling the result | 09:25.28 |
| | Clearly that is returning incorrect results 'sometimes' | 09:25.40 |
| CanadianHusky | so changing the color operators alone is not enough ? I see | 09:25.52 |
| kens | I won't know why until I have teh time to lok at the problem | 09:25.53 |
| | No, in general shadings are defined by a function (or an array of functions), in order to change the colour space we need to run the functions at certain values, record the results, convert the colour to gray scale, and then produce a new function which generates those results from the same inputs | 09:26.49 |
| CanadianHusky | have a nice weekend and thank you | 09:26.51 |
| aliray | hi, I have a pdf with black color. Is there a way to convert this black to 100% cmyk cyan? | 21:40.07 |
| | Forward 1 day (to 2018/11/11)>>> | |
Log of #ghostscript at irc.freenode.net.