[voiper1]

I recently went down the PDF rabbit hole for a project.

I had to use different OSS tools to do everything I wanted. I was able to access three from within nodejs without touching the disk:

1) Libreoffice CLI for converting doc/docx to PDF. It handled the formatting remarkably well. WARNING: you must have the fonts on the system doing the generating or it will substitute "similar" fonts! NPM: libreoffice-convert

2) NPM pdfjs-dist from mozila for extracting text and finding page numbers.

3) NPM pdf-lib for manipulating PDFs: deleting pages, adding pages from other PDF files (even to the middle of a PDF.)

4) PDF Jam commandline for resizing a pdf `pdfjam --keepinfo --outfile "${path}.resized.pdf" --paper letterpaper "${path}"`;

[danielrhodes]

Libreoffice only does a mediocre job of rendering Word documents. There are a number of cases where it really mangles things. An example would be some types of bulleted lists or indentation.

[wolverine876]

Given the GP's and your differing experiences, I wonder in which circumstances it works and in which it doesn't.

[danielrhodes]

In most cases it is adequate. But there are a couple factors to consider:

1) OpenXML is an open standard and like HTML it is interpreted and rendered, much like a browser. MS Word is obviously the reference here. But in certain cases, you will see differences when using other renderers. If you wrote the document in Word and then view it in LibreOffice or wherever, those differences are going to seem pretty glaring. It's possible to fix your document to not run into these issues, but because OpenXML has cascading styles and Word can often times produce quite messy output, it's hard to know what is going to not work well. It has taken the web 20+ years to get to a place where the difference between renderers is small enough to not be a big headache.

2) In OpenXML there is no concept of a page. Everything is relative and only at render time do you know how it looks. PDFs do have a concept of a page and are absolute - so that translation can be quite important and another source of rendering issues. An example of this is in Excel: people do need to print spreadsheets and if you want it to not look obnoxious you have to fiddle with the settings to get everything looking good. If you convert a spreadsheet to PDF you're more or less doing the same thing. However, in an automated context you can't make the judgements needed to make it look good - so you will often times end up with PDFs that have hundreds and hundreds of pages and look awful.

For 1) Microsoft could just release an API or some kind of package that gives you the output that matches Word and it would solve all of this. But at that point OpenXML ceases to be really open because nobody would want to use anything else. For 2) That's a lot harder to solve.

[wolverine876]

Thank you!