Two column layout not parsed correctly

[irian-codes]

Describe the bug Testing a two column layout PDF it seems the text gets misplaced in the final markdown result.

To Reproduce Parse Root board board game manual pdf or Aliens AGDITC board game manual

Expected behavior It should detect columns and join the text correctly in the final output without any other character in between. Right now, if you try to see the continuation of a left column to the right column in any of the example pdfs the text parts don't match.

For example, in the Root Manual:

• Text on page 8 left column on the PDF ends with Items on the Vagabond’s faction board can be face up or and continues on the right column with face down. The Vagabond exhausts face-up undamaged items, flipping them face down, to take many actions.. However, in the pdf2md output the text ends with 9.3.3 Step 3: Get Quests. which is the right column of the next page.

Example in the Aliens manual:

• Text on page 9 left column on the PDF ends withAfter placing the Blip tokens, read the game text and complete its instructions. and continues on the right column with You can get a sense of the danger coming by looking at the back of the Motion Tracker cards. However, in the pdf2md output the text ends with BLIPS \n GAME-TEXT \n TYPE LOCATION \n TUNNEL CARD: These cards place Tunnel tokens on the game board. which is the right column of the correct page but for whatever reason it entirely skips the top right column paragraph.

Desktop (please complete the following information):

• OS: Windows 11 running WSL
• Browser Node.js 22.6.0. (in a Next.js 14 app)
• Package version: "@opendocsg/pdf2md": "^0.1.31"

[irian-codes]

Okay seems columns are switched, like, it parses first the right column and then the left one. Which is incorrect for LTR documents.