Choosing Native PDF Output in pdfTeX: A Practical Decision Guide

Decision: Native PDF vs. PostScript Pipeline

When you need a PDF from a TeX source, you can either let pdfTeX write the PDF directly or route the DVI through dvips and pstopdf. The choice affects build time, compatibility, and the ability to embed advanced PDF features.

Constraints and Goals

• Target format: PDF only (no PostScript output required).
• Need to embed PDF specials (e.g., custom annotations or transparency).
• Prefer a single‑pass build to reduce build time in continuous‑integration pipelines.
• Maintain compatibility with older TeX installations that may lack native PDF support.

Options Overview

Engine	Output Path	Required Packages	Special Features	Compatibility	Typical Use‑Case
pdfTeX (native)	Direct PDF	None (built‑in)	Page‑level \\special{pdf: ...} for transparency, forms, etc.	All recent TeX Live / MiKTeX (≥ 2008)	Fast builds, advanced PDF features, minimal external tools.
dvips + pstopdf	PostScript → PDF	None (requires dvips and pstopdf)	Only what PostScript offers; limited PDF specials.	Older TeX distributions or when PostScript output is also needed.	Legacy workflows, when PDF specials are not required.
XeTeX	Direct PDF (via --pdf)	None (built‑in)	Unicode, OpenType font handling; limited PDF specials compared to pdfTeX.	Modern TeX Live (≥ 2016)	Unicode‑heavy documents, system fonts.
LuaTeX	Direct PDF (via --pdf)	None (built‑in)	Full Lua scripting; supports pdf.* primitives.	Modern TeX Live (≥ 2016)	Complex layout scripting, custom PDF manipulation.

Trade‑offs

• Build Speed: Native pdfTeX is one pass; dvips/pstopdf adds two passes.
• Feature Set: pdfTeX specials give fine control over PDF objects; XeTeX/LuaTeX have similar primitives but differ in syntax.
• Compatibility: Native pdfTeX works on almost all TeX installations; dvips/pstopdf is required on very old systems.
• Toolchain Complexity: Relying on external tools increases maintenance (e.g., ensuring dvips and pstopdf are in PATH).
• Debugging: Errors in specials can corrupt the PDF; careful catcode handling is needed.

Concrete Implementation

Below is a minimal example that demonstrates native PDF output and embeds a PDF form field using \\special{pdf: ...}. The example also shows how to verify that pdfTeX produced a PDF directly.

\\documentclass{article}
\\pdfoutput=1 % Enable native PDF output
\\begin{document}
Hello, native PDF!

% Embed a PDF textbox field named "name"
\\special{pdf: beginobj 1 0 R}
\\special{pdf: << /Type /Annot /Subtype /Widget /FT /Tx /T (name) /Rect [100 700 300 720] >>}
\\special{pdf: endobj}
\\end{document}

Run the build from a shell:

pdflatex -interaction=nonstopmode test.tex

Verification steps:

1. File header: Open test.pdf in a hex editor or use head -c 10 test.pdf; the first line should read %PDF- indicating a PDF file.
2. Producer field: Use pdfinfo test.pdf | grep Producer; the output should contain Producer: pdfTeX (or a similar pdfTeX identifier).
3. Log inspection: In test.log, look for lines like pdfTeX (pdf output) 3.14159265 and PDF code generation. Absence of such lines may mean the build fell back to PostScript.

Note: If you see a dvips or pstopdf invocation in the log, you are not using native PDF output.

Limitations and Practical Checks

• Specials are catcode‑sensitive; if you load packages that change catcodes around your \\special lines, the PDF code may break. Wrap specials in a \\begingroup ... \\endgroup to localise catcode changes.
• Not all PDF features are supported via specials; for complex forms or JavaScript, consider using the hyperref package with pdftex driver.
• Older pdfTeX releases (<2008) may not support \\pdfminorname and \\pdfmajorname primitives, affecting scripts that query the engine version.
• When cross‑platform builds are required, test on the target OS to confirm that the native output path is available.

Summary

If your workflow requires fast, single‑pass PDF generation and you need to embed low‑level PDF objects, enabling \\pdfoutput=1 in pdfTeX is the most straightforward choice. The dvips/pstopdf pipeline adds unnecessary complexity unless you must produce PostScript or are constrained to an old TeX distribution. XeTeX and LuaTeX offer native PDF output with different feature sets; choose them when Unicode or scripting is a priority. Always validate the resulting PDF header and Producer field to confirm that you are indeed using pdfTeX’s native PDF engine.