cv-sections-core.mjs
Rama main · réplica de lectura del repositorio original.
// Shared optional-section stripping for the CV builders (build-cv-html.mjs,
// build-cv-latex.mjs).
//
// Core competencies, work experience, projects, education, certifications,
// awards, interests, and skills are the genuinely optional CV sections: a
// competency tag row is often redundant with the summary and experience
// bullets that prove the same claims, a student or career-switcher has no
// professional history to list yet, a candidate's projects are often already
// covered under Work Experience, not every candidate has a degree, not every
// application carries a certification worth listing, most candidates have no
// award to name, most candidates don't add a personal-interests line, and
// plenty of candidates list no skills section at all. The templates wrap all
// eight unconditionally, so a payload with no entries renders a bare section
// header with nothing under it. The builders' buildCompetencies()/
// buildExperience()/buildProjects()/buildEducation()/buildCertifications()/
// buildAwards()/buildInterests()/buildSkills() correctly return '' — nothing
// removes the surrounding wrapper, which is what this module does.
//
// Work experience (#2504) is optional for the people this tool is aimed at —
// new graduates, career changers, and anyone leading with projects or
// education — not because the section is unimportant. Note that the payload
// key being optional says nothing about the template placeholder:
// cv-templates.mjs still requires `{{EXPERIENCE}}` to be present in any custom
// template, exactly as before. What changed is that an empty `experience` no
// longer leaves the wrapper behind.
//
// Certifications and Interests have no marker in the LaTeX template
// (cv-template.tex has neither section at all), so PATTERNS.tex has no
// `certifications`/`interests` key — stripEmptySections skips a section
// silently when the active format has no pattern for it, rather than trying
// to match against `undefined`. Awards, by contrast, is defined for both
// formats.
//
// The professional summary (#5019) is optional too: some house styles leave it
// out and open on the experience bullets. It is the one section held as text
// rather than as a list of entries, so isEmptySection() treats blank and
// whitespace-only text as empty. The LaTeX template has no summary, so it is
// html-only. It is also the first section in every shipped HTML template, so
// it uses the marker-only boundary described for Skills below. In a template
// with no marker after the summary, an end-of-input fallback would take the
// whole CV with it.
//
// ── The Skills sentinel: part of the template contract ───────────────────────
//
// Skills is the LAST section in every shipped template, which makes it the one
// optional section that may have no following section marker to stop at. Given
// the shared `…|$` boundary the others use, stripping an empty trailing Skills
// section falls through to true end-of-file and takes the closing
// `</div></body></html>` (`\end{document}` in LaTeX) with it — a truncated,
// unopenable document, which is far worse than the bare header this module
// exists to remove.
//
// So the Skills patterns below deliberately do NOT use the shared boundary.
// They use the same marker shapes with the `|$` branch removed, so they stop at
// the next marker and never at end-of-input. Because `END` is itself a marker,
// that is the `<!-- END -->` (`%%%% END %%%%` in LaTeX) sentinel when Skills is
// last, and the following section's marker when a custom template puts Skills
// somewhere else. Matching the sentinel *only* would be wrong for that second
// case: the lazy body would run past every section between Skills and the
// sentinel and delete them along with the empty header, which is silent data
// loss in a populated CV. Two consequences, both intentional:
//
// 1. **The sentinel is part of the template contract.** A template that ends
// on its Skills section must place `<!-- END -->` / `%%%% END %%%%`
// immediately after it. All four shipped templates do; do not remove it
// when editing a template's tail. This is documented for custom-template
// authors in templates/README.md.
// 2. **A template with no marker after Skills FAILS SAFE.** The pattern
// simply does not match, `String.replace` is a no-op, and the template
// comes out untouched — the Skills section renders as a bare header. That is
// the original cosmetic bug, and it is the deliberate choice: a bare
// header beats a truncated CV by a wide margin, and the person it lands
// on (a third-party template pack with no sentinel and no skills listed)
// did nothing wrong. cv-templates.mjs validates custom templates against
// `required: ['NAME', 'EXPERIENCE', 'EDUCATION']` and does not — and
// need not — require the sentinel, precisely because its absence is
// survivable. Never "fix" this by giving the Skills patterns an `|$`
// fallback; that trades a cosmetic bug for a destructive one.
//
// The section body is delimited by markers rather than parsed, so the boundary
// pattern carries the whole correctness burden and is easy to get subtly wrong:
//
// - Stopping at any capitalized comment would also stop at an ordinary
// comment inside a section body, truncating the strip and leaving markup
// behind. Markers are therefore matched as all-caps only.
// - Omitting the end-of-input branch would silently keep a section that
// happens to be last in the template. (Skills is the deliberate exception
// above — for it, keeping the section is the desired fail-safe.)
// - Naming the expected successor ("projects is followed by education")
// couples the two strips to each other and to template ordering: once an
// empty education block is removed, a named lookahead for it stops matching
// and the projects header survives.
//
// Each of those failure modes reintroduces the bare header this module exists
// to remove, and does it silently, so they are covered in
// tests/cv-optional-sections.test.mjs.
// HTML: `<!-- SECTION NAME -->`, all-caps. LaTeX: `%%%% Name %%%%` banners.
const HTML_BOUNDARY = String.raw`(?=<!--\s+[A-Z][A-Z ]*-->|$)`;
const TEX_BOUNDARY = String.raw`(?=%{4,}\s|$)`;
// Marker-only boundaries for Skills — the same marker shapes as the shared
// boundaries above, but with NO end-of-input alternative, so a template lacking
// any following marker is left untouched rather than truncated. `END` is itself
// a marker, so these stop at the `<!-- END -->` / `%%%% END %%%%` sentinel when
// Skills is last and at the next section's marker when it is not. See "The
// Skills sentinel" above before changing these, and never add an `|$` branch.
// The HTML summary uses the same marker-only shape for the same reason: it sits
// first, so a template with no marker after it must keep its bare header. It
// differs in one respect, HTML_SECTION_ONLY below, which does not stop at END.
//
// The LaTeX one anchors the banner to the start of a line (hence the `m` flag on
// the pattern that uses it). Without `^`, dropping the `|$` branch lets the
// engine backtrack the opening banner's own greedy trailing `%{4,}`: with no
// following banner to stop at it gives back `%` until the leftovers themselves
// satisfy the lookahead, matching half the banner and leaving a stray `%%%%`
// behind instead of no-opping. Only banners wider than 8 `%` can backtrack that
// far, so a narrow fixture will not catch a regression here — the fixtures in
// tests/cv-optional-sections.test.mjs use the shipped 28-wide banner on purpose.
const HTML_END_SENTINEL = String.raw`(?=<!--\s+[A-Z][A-Z ]*-->)`;
// The summary's own boundary: a following SECTION marker only. `<!-- END -->`
// is a marker but not a section, and a template may put it after an unmarked
// {{EXPERIENCE}}, so stopping there would strip the experience with the summary.
const HTML_SECTION_ONLY = String.raw`(?=<!--\s+(?!END\s*-->)[A-Z][A-Z ]*-->)`;
const TEX_END_SENTINEL = String.raw`(?=^%{4,}\s)`;
const PATTERNS = {
html: {
summary: new RegExp(String.raw`<!--\s+PROFESSIONAL SUMMARY\s+-->[\s\S]*?` + HTML_SECTION_ONLY),
competencies: new RegExp(String.raw`<!--\s+CORE COMPETENCIES\s+-->[\s\S]*?` + HTML_BOUNDARY),
experience: new RegExp(String.raw`<!--\s+WORK EXPERIENCE\s+-->[\s\S]*?` + HTML_BOUNDARY),
projects: new RegExp(String.raw`<!--\s+PROJECTS\s+-->[\s\S]*?` + HTML_BOUNDARY),
education: new RegExp(String.raw`<!--\s+EDUCATION\s+-->[\s\S]*?` + HTML_BOUNDARY),
certifications: new RegExp(String.raw`<!--\s+CERTIFICATIONS\s+-->[\s\S]*?` + HTML_BOUNDARY),
awards: new RegExp(String.raw`<!--\s+AWARDS\s+-->[\s\S]*?` + HTML_BOUNDARY),
skills: new RegExp(String.raw`<!--\s+SKILLS\s+-->[\s\S]*?` + HTML_END_SENTINEL),
interests: new RegExp(String.raw`<!--\s+INTERESTS\s+-->[\s\S]*?` + HTML_BOUNDARY),
},
tex: {
// The LaTeX banner is `%%%% Experience %%%%` — mixed case, and not the
// "WORK EXPERIENCE" the HTML templates use.
experience: new RegExp(String.raw`%{4,}\s+Experience\s+%{4,}[\s\S]*?` + TEX_BOUNDARY),
projects: new RegExp(String.raw`%{4,}\s+PROJECTS\s+%{4,}[\s\S]*?` + TEX_BOUNDARY),
education: new RegExp(String.raw`%{4,}\s+Education\s+%{4,}[\s\S]*?` + TEX_BOUNDARY),
awards: new RegExp(String.raw`%{4,}\s+AWARDS\s+%{4,}[\s\S]*?` + TEX_BOUNDARY),
skills: new RegExp(String.raw`%{4,}\s+Technical Skills\s+%{4,}[\s\S]*?` + TEX_END_SENTINEL, 'm'),
},
};
export const OPTIONAL_SECTIONS = ['summary', 'competencies', 'experience', 'projects', 'education', 'certifications', 'awards', 'interests', 'skills'];
// Sections whose payload value is text rather than a list of entries.
const TEXT_SECTIONS = new Set(['summary']);
// A list is empty when it has no entries. stripEmptyRenderedSections() hands
// every section in as a list, text sections included, so that check runs first.
// Text is empty when it would render nothing: absent, blank, whitespace-only,
// or a structural value escapeHtml() prints as ''. A number still renders, 0
// included, so it counts as content.
export function isEmptySection(payload, section) {
const value = payload?.[section];
if (Array.isArray(value)) return value.length === 0;
if (TEXT_SECTIONS.has(section)) return value == null || typeof value === 'object' || String(value).trim() === '';
return true;
}
// Remove every optional section that has no entries in `payload`. Returns the
// template unchanged when both are populated.
export function stripEmptySections(template, payload, format) {
const patterns = PATTERNS[format];
if (!patterns) throw new Error(`Unknown template format: ${format}`);
let out = template;
for (const section of OPTIONAL_SECTIONS) {
const pattern = patterns[section];
if (!pattern) continue; // this format's template has no marker for this section
if (isEmptySection(payload, section)) {
// A non-matching pattern is a no-op here by design — see the Skills
// sentinel note above: no sentinel means no strip, never a fallback to
// a looser boundary.
out = out.replace(pattern, '');
}
}
return out;
}
// ── Content-driven emptiness, for callers holding rendered HTML ──────────────
//
// isEmptySection answers from the payload the template was filled from. The
// web pdf path has no payload to ask: since #2185 the agent emits finished
// HTML and the backend writes it straight to disk, so neither builder — and
// therefore neither stripEmptySections call site — is on that path, and every
// optional section the agent left empty reaches the renderer as a bare header.
//
// The two exports below answer the same question from the rendered document
// instead. They read the section body back out with the PATTERNS.html entry
// that stripEmptySections removes with, so a marker or boundary change moves
// both halves together and there is still one definition of where a section
// starts and ends.
// A section's own title is present whether or not the section has content, so
// it must not count as content. This matches the shipped templates'
// `<div class="section-title">`, or a plain heading tag for a custom template
// that uses one. Either quote style, matching generate-pdf.mjs's own
// SECTION_TITLE_RE: a hand-written or third-party template may use single
// quotes, and reading its title as content would keep every empty section.
// Applied once — a later heading inside a populated body
// keeps its text and correctly reads as content.
const SECTION_HEADING_RE = new RegExp(
String.raw`<(div|h[1-6])\b[^>]*class=["'][^"']*\bsection-title\b[^"']*["'][^>]*>[\s\S]*?<\/\1>` +
String.raw`|<(h[1-6])\b[^>]*>[\s\S]*?<\/\2>`,
'i',
);
// Does this section render as a bare header — a title and nothing under it?
//
// False when the format has no pattern for the section, and false when the
// document has no such section at all: both mean there is nothing to strip,
// and reporting "empty" for a section that is simply absent would invite a
// caller to act on markup that was never there.
export function isEmptyRenderedSection(html, section) {
const pattern = PATTERNS.html[section];
if (!pattern) return false;
const block = html.match(pattern);
if (!block) return false;
const text = block[0]
.replace(/<!--[\s\S]*?-->/g, ' ')
.replace(SECTION_HEADING_RE, ' ')
.replace(/<[^>]*>/g, ' ')
.replace(/ |\u00a0/g, ' ');
return text.trim() === '';
}
// Remove every optional section that rendered as a bare header. The emptiness
// map is derived from the document as it arrives, before anything is removed,
// so one strip cannot change the verdict on the next.
export function stripEmptyRenderedSections(html) {
const derived = {};
for (const section of OPTIONAL_SECTIONS) {
derived[section] = isEmptyRenderedSection(html, section) ? [] : [section];
}
return stripEmptySections(html, derived, 'html');
}