In October I ported ReadableResume's Classic resume template for Google Docs to LaTeX, Quarto and Zola. The LaTeX port became atsresume, a LaTeX class on CTAN. The Quarto port renders through Typst, and its source and a sample resume are on GitHub. For HTML I wrote readable-resume, a one-page theme in Zola's theme gallery.
The layout has a centered name, uppercase headings over a thin rule, and the city and dates flush right. In one Zola test print, pypdf put all 12 bullets after the Skills section, even though the page looked fine.
Write the numbers down once
The values come from a PDF export of the Docs file: Georgia at 10.5 pt with 1.1 line spacing, a 20 pt name, 11 pt headings over a 0.75 pt rule, and margins of 0.6 in on US Letter or 15 mm on A4. In Georgia, 1.1 line spacing works out to lines about 13.1 pt apart. The three ports spell that number differently:
% LaTeX: font size and baseline-to-baseline distance, in points \renewcommand\normalsize{\@setfontsize\normalsize{10.5}{13.1}} // Typst: the gap from one line's baseline to the next line's cap height #let _leading = 0.55em /* CSS: 10.5pt x 1.25 = 13.125pt */ body { font: 10.5pt/1.25 var(--font); } Typst is the odd one. Its leading isn't the distance between baselines. By default a line's box runs from the cap height down to the baseline, and leading is the space between those boxes. Subtract Georgia's cap height, about 0.69 em, from the 1.25 em line and you get roughly 0.56 em. The template uses 0.55 em.
The first line needed the same kind of fix. Google Docs puts the first baseline one ascent below the top margin, and Typst puts it one cap height below. The Typst template adds a top inset of a quarter of the name's size to make up the difference, and atsresume puts an invisible 18.3 pt strut on the name line.
Fonts couldn't match everywhere. Georgia isn't in TeX Live, so atsresume uses TeX Gyre Pagella, whose letters are about as wide as Georgia's. Many lines wrap in the same places, and the sample resume compiles to one page. A resume that already fills the page in Docs can still wrap differently. The class loads the Type 1 version under pdfLaTeX and the OpenType version under LuaLaTeX and XeLaTeX. The Zola theme's CSS asks for Georgia, then Gelasio, then Times New Roman.
The right-aligned city and dates
A table or a second column would make the right side easy, but PDF extractors can then read the city apart from the employer. All three ports put the employer before the city in the source, with the city and dates flush right. If the right part won't fit beside the text on the left, it drops to the next line and stays flush right.
In LaTeX it's glue and penalties, a trick from chapter 14 of The TeXbook:
\rightskip\z@\@plus 4em\relax \noindent#1% \unskip\nobreak\hfil\penalty50\hskip1em\hbox{}\nobreak\hfil \mbox{#2{#3}}% \parfillskip\z@skip \penalty50 is where TeX may break between the two parts. If both parts fit, the two \hfil share the gap and the right part ends at the margin. If not, TeX breaks there: the first \hfil fills out the left line, and the empty \hbox{} keeps the second \hfil from being thrown away at the start of the new line. \rightskip gets only finite stretch, so a long company name still wraps ragged while the infinite \hfil decides where the date goes.
Typst does it in one line:
#let _row(left, right) = { if _filled(left) { left } if _filled(right) [ #h(1fr)#sym.wj#box(right)] } The space before #h(1fr) is where the line can break. box() keeps a date range in one piece, and the word joiner sym.wj stops a break from landing between the spacing and the box.
In CSS it's a wrapping flex row:
.row { display: flex; flex-wrap: wrap; column-gap: 1em; } .right { margin-left: auto; white-space: nowrap; } The HTML puts the employer before the city. In the sample PDF printed by Chrome, pypdf returned Harlan Quay Capital Partners New York, NY on one line.
Bullets can reorder the PDF text
One common way to hang a custom bullet in HTML is an absolutely positioned ::before. The Zola theme avoids that. Its stylesheet warns that bullets positioned this way can turn up later in the text extracted from Chrome's printed PDF.
I wanted to see it, so I built the theme's sample site with Zola 0.23.6 and printed it with headless Chrome 157 twice. The first print used the theme's CSS. The second added this override:
li { position: relative; text-indent: 0; } li::before { position: absolute; left: 0; top: 0; } The two pages looked identical. pypdf 7.1.1 read them very differently. From the override, every heading and job line came first, then the Skills section, and only after that all 12 bullets:
Halden Habitat Builders | Volunteer Coordinator Sep 2022 – May 2024 SKILLS Technical: Excel, PowerPoint, Python (pandas), Stata, Capital IQ Certifications: Bloomberg Market Concepts, Wall Street Prep Financial Modeling Languages: French (fluent), Spanish (basic) GPA: 3.8/4.0, Economics Department Honors Program• A parser that splits the pypdf output at section headings would put all 12 bullets under Skills. pdfminer.six kept the bullet text near its entries but put the dots in separate blocks.
The theme's version uses an inline-block dot and a negative indent, so nothing is positioned:
li { padding-left: 0.15in; text-indent: -0.15in; } li::before { content: "•"; display: inline-block; width: 0.15in; text-indent: 0; } From that PDF, pypdf and pdfminer.six both returned all 12 bullets starting with •, each under its own entry.
The LaTeX class has its own note on bullets. It prints the text font's U+2022 at full size on the baseline. According to the source, a smaller, raised dot looks closer to the Docs original, but pdftotext and other readers that guess layout can split such dots off as if they were a column. The Typst template does the same thing with marker: box(width: 0.15in, align(start)[•]).
Contact links
Email addresses and URLs needed different handling in each engine.
LaTeX. An underscore in an email address is the classic failure. With plain hyperref,
\href{mailto:first_last@example.com}{first_last@example.com} pdfLaTeX stops with ! Missing $ inserted., because _ in the link text means a subscript. atsresume's \email and \weblink read their argument as a plain string with expl3. They drop escapes like \_ that people type out of habit, use the string as the link, and then rescan it for printing so that under pdfLaTeX a UTF-8 letter prints as that letter. This compiled under both pdfLaTeX and LuaLaTeX:
\contact{\email{zoë.first_last@example.com}} \contact{\weblink{github.com/first_last}} \contact{\weblink{https://example.com/a#b}} The text layer had all three addresses exactly as typed, and the links pointed to https://github.com/first_last and https://example.com/a#b. A bare domain gets https:// in the link only.
Quarto and Typst. Quarto's Pandoc writes Typst markup, and in Typst markup @name is a reference, so the writer escapes it:
$ printf 'olivia.bennett@example.com\n' | quarto pandoc -f markdown -t typst olivia.bennett\@example.com That's correct in markup and wrong in a string. The template builds the contact links from strings, and Typst 0.15.1 doesn't treat \@ as an escape inside one. It keeps the backslash, so link("mailto:" + item, item) printed olivia.bennett\@example.com, and the PDF link pointed to mailto:olivia.bennett\@example.com. The extension's Lua filter therefore turns the name and contact items into plain text and writes the Typst strings itself, escaping only backslashes and quotes:
local function typst_string(text) local escaped = text:gsub("\\", "\\\\"):gsub('"', '\\"') return '"' .. escaped .. '"' end Two Typst syntax details came up while writing that filter. A one-item list has to end with a comma, because ("a") is a string in parentheses and ("a",) is an array. An empty list can't be written as (,), which fails with "unexpected comma", so the filter drops the field and the template's default applies.
There's one more gotcha on the Quarto side. Quarto reads typst-template.typ as a Pandoc template partial, where $ marks a variable, so the $ at the end of the template's URL regex is written $$. When I changed it to a single $, the render stopped with Error compiling template ... unexpected "\"". Quarto also numbers Typst pages by default, so the extension sets page-numbering: false.
Zola. Contact items in the front matter are strings or { text, url } tables. A string becomes a link if it's a bare email address or starts with http://, https:// or mailto:. Other strings go through Zola's inline Markdown filter, so a bare linkedin.com/in/... stays plain text. To make it clickable, write a Markdown link or give the item a url:
{ text = "github.com/oliviabennett", url = "https://github.com/oliviabennett" } A nonempty url in a contact table must start with https://, http://, mailto: or tel:. Any other scheme, such as a pasted javascript: link, stops the build. Links to other sites get the same rel attributes that [markdown] in zola.toml gives Markdown links. With the same sample contact line, the Chrome PDF had one link (the email), while the Quarto PDF had two (the email and https://linkedin.com/in/oliviabennett).
Typos in the resume data
Tera, Zola's template engine, has no schema for these resume fields. The template reads location with default(value=""), so without a field check, typing locaton would just leave the city blank. So the template checks the keys of every section, entry, contact table and labeled line against a list and calls throw() on anything else. With the check in place, the same typo makes zola build exit with status 1:
ERROR Reason: error: content/_index.md: unknown field "locaton" in an entry under "Experience". Expected one of: company, location, title, date, bullets. resume_paper in zola.toml gets a separate check. It has to be letter or a4, and the template uses it to write the @page size and margins and the CSS variables that draw the sheet on screen. In LaTeX, \makeheader without a \name stops with Class atsresume Error: No name given.
What I'd do on the next port
Next time I'd keep the Google Docs measurements in one place first, so values like Typst's 0.55 em leading can be worked out from them. Before comparing screenshots, I'd run the sample PDF through pypdf and pdfminer.six and check that the GPA bullet still comes under Education.
I work on ReadableResume, a free resume template site, and these three ports are ours. atsresume is under the LaTeX Project Public License 1.3c. The Quarto extension and the Zola theme are MIT-licensed.
I used AI to help draft this post.
