Release Notes

Version 0.5.2

Changes

Added SYNTAX.md, a one-page syntax reference: every keyword with where it is allowed, stored values, condition rules, load-time validation errors, and known limits (features that parse but don't render yet). Its survey examples are parsed by tests/syntax-examples.test.ts so the reference stays in sync with the parser

Bug Fixes

Fixed answers to hidden questions still steering conditions: changing an earlier answer so that a later, already answered question became hidden (e.g. unticking a checkbox option, or picking a different radio option) left that later answer in the variables, so it kept driving SHOW_IF, COMPUTE, and {{IF}}/{var} placeholders even though the respondent could no longer see or change it. Variables are now derived only from content that is visible: answers inside a hidden block, page, section, question, or matrix row, and selected options hidden by their own SHOW_IF, count as unanswered. The answers themselves are kept, so they count again as soon as the content becomes visible. Because visibility depends on the variables, derivation repeats until the result is stable, which resolves chains where hiding one question hides the next
- deriveVariables and useQuestionnaireResponses now take the survey's blocks instead of a flat page list, so block-level SHOW_IF and computed variables can be evaluated

Internal

Updated dependencies: Next 16.3.4 → 16.3.6, React 19.2.8 → 19.3.0, lucide-react 1.43 → 1.48, tailwind-merge 3.6 → 3.7, ESLint 10.10 → 10.11, and Vitest 4.1 → 5.0 (no test changes needed). TypeScript stays on 6.0 because typescript-eslint (via eslint-config-next) does not support 7 yet

Version 0.5.1

Changes

Added a GitHub link to the navbar: an icon-only button linking to the project repository, shown on the home, documentation, and releases pages

Bug Fixes

Fixed checkbox/radio - TEXT options corrupting their stored value and breaking SHOW_IF: typed text was encoded directly into the response string (e.g. "Other, namely:" + ": " + "painting"), which produced a doubled separator whenever the option's own label already ended in punctuation ("Other, namely:: painting"), and made SHOW_IF/== comparisons against the bare option value silently stop matching as soon as any text was typed. Typed text is now stored separately from the selected value, so responses/{variable} always hold the plain option value(s) selected — conditions keep matching regardless of typed text — while {variable AS LIST}/{variable AS INLINE_LIST} and other text placeholders join the option's label with its typed text using a single space ("Other, namely: painting"), so the label's own punctuation controls how they read together instead of the app inventing a separator
Fixed checkbox response order following click order instead of option order: a checkbox question's stored array (and therefore {variable AS LIST}) reflected the order options were checked in, not the order they're declared in the survey. Selections are now always ordered to match the question's option list

Internal

Linting works again: pnpm lint ran next lint, a command Next 16 removed, and eslint-config-next had been dropped from devDependencies by an earlier dead-code pass (it is referenced by the ESLint config rather than imported, so the tool could not see the reference). The script now runs eslint ., eslint-config-next is back, and eslint.config.mjs uses its native flat configs directly instead of the FlatCompat shim (@eslint/eslintrc is no longer needed). eslint-plugin-react's React-version autodetection calls an API removed in ESLint 10, so the config pins settings.react.version; @next/next/no-img-element is off because markdown images come from arbitrary survey text
Removed setState-in-effect patterns flagged by the React Compiler lint rules: eight effects that synchronised state React can derive or read directly were rewritten, so no component renders once with a placeholder value and then again with the real one
- Page index bounds are clamped during render (clampPageIndex) instead of corrected afterwards, so an out-of-range index is never rendered when answers hide pages. A respondent whose current page becomes hidden is moved to the last visible page and returns to where they were if those pages become reachable again
- Visited-page tracking moved from RespondentNavigator into useQuestionnaireNavigation, which now takes the visible pages and returns the set of page ids reached. The navigator derives visited nav items from it rather than accumulating them in an effect
- Expansion state in both navigators is now derived — the page navigator expands the block holding the current page, the respondent navigator expands groups whose pages have been reached — with manual open/close choices remembered until the next navigation, matching the previous reset-on-navigate behaviour
- localStorage preferences (language, navigation settings) are read through useSyncExternalStore (hooks/use-local-storage.ts) so the stored value is applied as part of hydration, and the survey draft editor reads its draft in a lazy useState initialiser since it only mounts after the user opens it
- Mac detection for the keyboard-shortcut hint reads navigator through the same store pattern instead of a mount effect
Switched the package manager to pnpm: package-lock.json is replaced by pnpm-lock.yaml, package.json pins packageManager, and the npm-only overrides field moved to overrides in a new pnpm-workspace.yaml (pnpm 11 no longer reads settings from package.json). Re-resolving the dependency tree from the version ranges also picked up newer patch/minor releases (Next 16.2.9 → 16.3.4, React 19.2.7 → 19.2.8, lucide-react 1.17 → 1.43, TypeScript 6.0.3, ESLint 10.10, Vitest 4.1.11)
- TEXT other-text is now lifted into useQuestionnaireResponses (an otherTexts map alongside responses) instead of living in local useState inside checkbox-question.tsx/radio-question.tsx. A new pure applyOtherText (lib/response-variables.ts) composes it into a separate displayVariables object for text rendering, while variables (used for SHOW_IF/expressions) stays plain. displayVariables/otherTexts are threaded alongside the existing variables/responses props through page-content.tsx → section-renderer.tsx → question-renderer.tsx → the question components; removed the now-redundant colon-parsing duplicated in checkbox-question.tsx, radio-question.tsx, and lib/utils/tab-index-calculator.ts

Version 0.5.0

Changes

Rewritten condition engine: SHOW_IF and COMPUTE expressions are now evaluated by a real expression parser instead of string splitting. What this changes for surveys:
- Parentheses work: SHOW_IF: (age >= 65 OR age < 30) AND consent == Yes now groups correctly (previously it silently mis-evaluated)
- Keywords are UPPERCASE-only: AND, OR, NOT, IS, IS_NOT, STARTS_WITH etc. are only recognized in uppercase, so unquoted answer values containing words like "or" ("Several weeks or more") are no longer torn apart. Lowercase and/or/not are now treated as plain text
- NOT binds tighter than AND/OR: NOT a AND b now means (NOT a) AND b (conventional precedence); previously it negated everything after it. No surveys or examples in this repository used the old interpretation
- String comparisons fixed: comparing two text variables (name1 == name2) now compares their text; previously both sides were coerced to numbers, making any two non-numeric values "equal"
- STARTS_WITH accepts quoted values: STARTS_WITH crime == "Yes" now works the same as the unquoted form
- No more JavaScript evaluation: arithmetic is evaluated by the parser itself instead of generated JavaScript code, so survey text can never execute code
Malformed conditions are now rejected when a survey is loaded: every SHOW_IF and COMPUTE expression — on blocks, pages, sections, questions, options, and matrix subquestions — is syntax-checked at parse time, and errors name the exact location (e.g. Question "Q3" option "Sometimes" SHOW_IF "mode == (": Unbalanced parentheses). Previously a broken condition silently kept its content visible; surveys that relied on that will now fail to load until the condition is fixed. The runtime evaluator keeps a fail-safe default (content stays visible plus a console warning) as a last resort for conditions that slip past validation
Variable validation is more thorough: duplicate variable names are now caught across all variable-defining locations — question, matrix subquestion, breakdown option, and computed variables — not just question-level VARIABLE:. SHOW_IF undefined-variable checks now also cover sections and matrix subquestions, matching the coverage questions and options already had
EXCLUSIVE checkbox options: A checkbox option can now be marked - EXCLUSIVE (indented under the option, like - TEXT) so selecting it deselects every other selected option, and selecting any other option deselects it. Useful for options like "None of the above". A question can have more than one exclusive option; selecting one always deselects the others. Exclusive options render with a circular, radio-style indicator to signal this to respondents
Multi-line Q: text: Question text can now span multiple lines using """ delimiters, the same convention already used by HINT:, REVEAL:, and TOOLTIP:. This lets a question reference a bulleted or numbered list of examples as part of its own text (e.g. "has your employer provided any of the following measures: ...") without those list items being mistaken for answer options
- Example: Q: """ on its own line, the markdown content (including a - or * list), then a closing """ on its own line, followed by the actual options
Summed comparisons in COMPUTE:: A computed variable can now add up multiple comparisons, e.g. COMPUTE: hazard_total = heat == Yes + cold == Yes + severe == Yes, counting 1 for each matching term. Plain numbers can be mixed into the sum, and the result can be used like any other computed variable (placeholders, SHOW_IF)
TOOLTIP: is now a popover, REVEAL: is the old inline panel: The previous TOOLTIP: keyword toggled an inline info panel below the element; that behavior is now REVEAL:. TOOLTIP: instead shows a small popover next to the text on click, for shorter contextual hints. Both are supported on pages, sections, questions, all option types (including multiple-choice and checkbox), and matrix subquestions, and can be combined
- Existing surveys using TOOLTIP: for inline panels should switch to REVEAL: to keep the same behavior
HINT:, TOOLTIP:, and REVEAL: now render on multiple-choice and checkbox options: these were already parsed on every option type but silently ignored for anything other than breakdown options. A - HINT:/- TOOLTIP:/- REVEAL: indented under a - option now shows the same muted subtext, popover, and collapsible panel as elsewhere
Tooltip and reveal icons now sit consistently next to the text everywhere: pages, sections, questions, breakdown/matrix/multiple-choice/checkbox options previously used two different layouts — a fixed left-margin icon for REVEAL: and an icon trailing the text for TOOLTIP: — which drifted apart once a label wrapped to multiple lines (the left-margin icon centers on the whole block; text-trailing follows the last line). Both icons now trail the text everywhere, in the same order (tooltip, then reveal), and are sized to fit inside a line of text instead of overhanging it. The page no longer reserves a left icon gutter (pl-8), since nothing is positioned there anymore
Empty checkbox variables and list conjunctions are now localized: {selections} on an unanswered checkbox variable rendered as the English word "none" and {colors AS INLINE_LIST} always joined the last item with "and", regardless of the survey's language setting. Both now follow the current language (Dutch: "geen"/"en")
Block-level computed variables are now global: A COMPUTE: defined at block level is visible everywhere in the survey, not just within its own block
- Block-level computeds can reference each other across blocks (resolved by dependency order)
- A SHOW_IF: on any block, page, section, or question can now reference a computed defined in any other block
- Page-level computeds remain page-scoped, seeded with the global block-level set
- Names must be unique across blocks — defining the same compute name in two different blocks is now a validation error
- Multiple COMPUTE: statements for the same name within a single block are still allowed (the default-then-override pattern)
Fixed documentation sidebar scrolling: The sidebar now scrolls independently on long pages instead of overflowing the viewport
Updated dependencies: Bumped all packages to latest versions; added a postcss override to resolve a XSS vulnerability (GHSA-qx2v-qp2m-jg93)

Bug Fixes

Fixed page navigator block visibility: The page navigator was re-evaluating block SHOW_IF conditions using the current page's computed variables instead of the global computed variables, which could cause it to show a different visibility state than the actual survey navigation. The navigator now derives block visibility directly from the viewer's own computation.
Hidden pages no longer clutter the nav sidebar: NAVIGATION items were built once at parse time and shown unconditionally, so a page excluded by SHOW_IF still appeared in the respondent's sidebar (permanently greyed out). Nav items are now filtered by current page visibility, and reappear automatically once the respondent's answers make the page reachable again. A visible level-2 item whose level-1 parent is hidden is promoted to a top-level entry so it isn't lost from the nav.
Fixed dash-bulleted lists inside """ blocks being parsed as answer options: A - line inside a delimited HINT:, REVEAL:, or TOOLTIP: block (e.g. a bullet list of examples) was being scraped by the option parser as a real, selectable option in addition to being rendered as text. Delimited content is now correctly excluded from option and question-type detection, so - and * lists inside any """ block render as plain lists and never become options
Fixed checkbox options rendering as circles: All checkbox inputs, not just EXCLUSIVE ones, were rendering as full circles instead of rounded squares. The checkbox's corner radius used the theme's --radius-lg (10px), which exceeds half the checkbox's 16px size — the browser clamps that to a perfect circle regardless of the intended shape. Non-exclusive checkboxes now use a fixed 4px radius so they render as squares again; EXCLUSIVE options are unaffected and still render as circles
Fixed matrix tables being too wide with few response options: The row-label column had a fixed 25% width, so with table-fixed layout the remaining 75% was split evenly across however many response-option columns existed — a 2-option matrix (e.g. Yes/No) ended up with very wide, mostly-empty columns. Response-option columns now use a fixed width instead, and the row-label column takes up the remaining space
Fixed question type detection rejecting options starting with "Q": A question whose options all began with the letter "Q" (e.g. "Quality", "Quantity") was parsed as a plain text question with no options — the option-detection check excluded any dash line starting with "Q" in an attempt to skip matrix subquestion markers (- Q1: ...). Option detection now shares the same logic used to actually build the option list, so only real subquestion markers and metadata lines are excluded
Fixed section REVEAL: content being silently dropped: Section-level REVEAL: text was parsed correctly but never reached the page — the visibility-filtering step rebuilt each section object without copying the reveal field over. The reveal button and panel are also no longer restricted to sections that have a title
Fixed bare page-level REVEAL:/TOOLTIP: swallowing later keywords: When a page used REVEAL: or TOOLTIP: with no text on the same line, any NAVIGATION:, COMPUTE:, or SHOW_IF: line that immediately followed was absorbed into the reveal/tooltip text instead of being processed — silently losing the page's nav level, computed variable, or visibility condition
Fixed the bundled sample survey: quoted the experienced_user COMPUTE comparison value and fixed a typo, so "Load Sample Survey" shows its intended conditional block
Fixed breakdown option SHOW_IF being silently ignored: - SHOW_IF: on a breakdown row was parsed and validated but never affected rendering — every row always showed regardless of the condition. Hidden rows are now excluded from the table and from totals/subtotals; their stored value is kept and restored if the row becomes visible again
Fixed order-dependent variable derivation: variables were derived by walking responses in the order the respondent answered them, not the order questions appear in the survey. A breakdown row's VALUE: placeholder referencing another question's variable could resolve differently (or not at all) depending on the order the two questions were answered, e.g. via back-navigation. Variables are now derived by walking the questionnaire's own page/section/question order, so identical answers always produce identical variables
Question-level VARIABLE: on a matrix question is now a parse error: matrix responses are stored per row, so a bare VARIABLE: at the question level (as opposed to - VARIABLE: under a row) was silently accepted but never populated — any condition, placeholder, or computed variable depending on it just silently never resolved. This now fails to load with a message pointing to the per-row - VARIABLE: syntax
Fixed forward-referencing CUSTOM: breakdown subtotals: a CUSTOM: subtotal formula could only resolve another subtotal's variable if that subtotal appeared above it in the option list — the value was written into the lookup map row-by-row as the table rendered, so a reference to a subtotal defined further down silently failed. Subtotals are now precomputed as a single, order-independent pass before any CUSTOM: formula is evaluated, so a CUSTOM: row can reference any other subtotal's variable regardless of table position. This also fixes the same latent issue in variable derivation (lib/response-variables.ts), which shared the identical row-by-row logic

Internal

Reduced code duplication across lib/parser.ts, lib/validation.ts, and components/questions/breakdown-question.tsx (~245 lines removed)
Consolidated variable-definition collection in lib/validation.ts into a single collectVariableDefinitions helper, reused by both the name-uniqueness and reference validators
Extracted a shared BaseOption type for the fields Option and BreakdownOption actually have in common (value, label, hint, reveal, tooltip, showIf), and fixed type-narrowing errors in parser-option-exclusive.test.ts/parser-option-text.test.ts surfaced by tsc --noEmit
Extracted breakdown row visibility/summing into lib/breakdown-calculations.ts and the two-pass variable derivation out of use-questionnaire-responses.ts into a pure, directly-testable lib/response-variables.ts; the hook is now a thin useState wrapper around it
Extracted breakdown subtotal computation (computeSubtotalValues/subtotalVariables) into lib/breakdown-calculations.ts, shared by breakdown-question.tsx and lib/response-variables.ts instead of being duplicated
Removed render-time mutation of shared parsed state: evaluateComputedValues no longer writes back onto ComputedVariable.value (the field is deleted from the type; nothing read it) — computed values are only ever returned, never stored on the parsed questionnaire
Deduplicated parser state machines in lib/parser.ts: a pushCurrentOption helper replaces the triplicated option-push block in parseOptions; shared beginMetadataCollection/stepMetadataCollection helpers (single-line/bare/"""-delimited, parameterized by a terminator predicate) replace the near-identical REVEAL/TOOLTIP/SHOW_IF state machines duplicated across parsePage and parseSection
Extracted breakdown-question.tsx's option label rendering into components/questions/shared/option-label-content.tsx, shared by radio-question.tsx and checkbox-question.tsx
Removed the gutter/absolute-positioning path for REVEAL: icons (question-header.tsx, page-header.tsx, section-renderer.tsx) in favor of the same text-trailing layout used by tooltips and options; deleted the now-unused pl-8/-ml-8 gutter compensation from questionnaire-viewer.tsx and components/ui/table.tsx

Version 0.4.0

Released May 2026

New Features

Conditional string values in COMPUTE: Computed variables can now return strings conditionally
- IF <condition> THEN <value> ELSE <value>: COMPUTE: label = IF score >= 8 THEN "High" ELSE "Low"
- ELSE IF chaining: COMPUTE: label = IF score >= 8 THEN "High" ELSE IF score >= 5 THEN "Medium" ELSE "Low"
- One-sided IF <condition> THEN <value> (no ELSE): leaves the variable unchanged if the condition is false, enabling a default-then-override pattern with multiple COMPUTE statements for the same variable
- Values can be quoted strings, numbers, or references to other variables
- The condition uses the same syntax as SHOW_IF
- String results can be used in text placeholders ({label}) and in SHOW_IF comparisons
String literal COMPUTE values: Assign a fixed string to a computed variable with COMPUTE: category = "high"
Single-line breaks in text: Single newlines in text content now render as line breaks, making it possible to place text on new lines without creating a new paragraph

Improvements

Package updates: Updated all dependencies to their latest versions

Bug Fixes

Breakdown option variables recognized in SHOW_IF: Variables defined on breakdown question options (via - VARIABLE:) are now correctly recognized during validation, so referencing them in SHOW_IF conditions no longer produces a false "undefined variable" error

Removed

Static survey pages: Removed all hardcoded survey routes (e.g. /r1-groothandel, /r8-bouw) along with their API routes and source files — these were added to share URLs that immediately loaded a specific survey, but are no longer needed

Version 0.3.1

Released January 2026

New Features

Section SHOW_IF support: Sections can now be conditionally shown or hidden based on variables and responses
- Uses the same SHOW_IF: syntax as pages and questions
- Hidden sections and their questions are excluded from page completion tracking
- Example: SHOW_IF: is_manager == Yes on a ## section heading
Section TOOLTIP documentation: Section-level tooltips are now documented with interactive examples
- The tooltip rendering was already implemented but undocumented — an info icon appears next to the section heading and toggles the tooltip content
- Uses the same TOOLTIP: syntax as pages and questions (including multi-line """ delimiters)
Page Navigator documentation: Added dedicated documentation page explaining the Page Navigator feature
- Covers overview, how to open it, and its features (page status indicators, jump-to-page, completion tracking)

Breaking Changes

Removed - ESSAY and - OTHER option modifiers: Only - TEXT is now supported for adding text inputs to options
- - ESSAY and - OTHER were functionally identical to - TEXT and have been removed to simplify the syntax
- Surveys using - ESSAY or - OTHER on options should replace them with - TEXT

Improvements

URL-based documentation routing: Each documentation section now has its own URL
- Refreshing the page stays on the current section instead of resetting to Overview
- Sections are bookmarkable and shareable (e.g., /docs/matrix, /docs/conditionals)
- Implemented as static Next.js routes with a shared layout
Polished documentation
- Went over the documentation pages and added clarifications where needed, improved consistency between pages, and improved styling
Documentation language override: Documentation examples now always render in English regardless of the user's language setting
- Added defaultLanguage prop to LanguageProvider to force a specific language
- Prevents Dutch placeholders and button labels from appearing in documentation
Expanded Overview page: Added detailed introductory text explaining what TST does and its design philosophy
Collapsible navigation sections: The chevron on expandable navigation items is now an independent click target
- Clicking the chevron toggles expand/collapse without navigating
- Clicking the item text still navigates (and auto-expands if collapsed)
- Previously, expanded sections could not be collapsed

Bug fixes

Documentation state bug: Fixed a bug where navigation between pages in one documentation example affected the navigation state in another documentation example
- Added key={activeSection} to force React to remount examples when navigating between sections
- Previously, navigating in one section's example could affect examples on other pages

Version 0.3.0

Released December 2025

Major Changes

Parser refactor: Complete rewrite using chunk-based hierarchical architecture
- Reduced parser code from ~2600 lines to ~1100 lines (60% reduction)
- Hierarchical parsing: Questionnaire → Blocks/NavItems → Pages → Sections → Questions
- Type-specific parsers for all 7 question types
- Eliminated indentation-based parsing - dash markers now identify metadata
- Improved maintainability and extensibility
Single-pass parsing architecture: Refactored parseSection, parsePage, and parseBlock to use state machines
- Fixed bug where multi-line delimited tooltip content was rendered twice (as metadata and as content)
- Each parser uses specific state names describing exactly what's being parsed (e.g., 'tooltip' | 'showif' | 'content' instead of vague 'normal')
- Structural markers (#, ##, BLOCK:) extracted immediately before entering state machine
- parseSection: Single-pass state machine extracts section metadata (TOOLTIP, SHOW_IF) and builds content/questions in one loop
- parsePage: Single-pass state machine extracts page metadata (NAVIGATION, TOOLTIP, COMPUTE) before building sections inline
- parseBlock: Single-pass state machine extracts block metadata (SHOW_IF, COMPUTE) before building pages inline
- Removed unused helper functions: identifySections, identifyPages, parseComputedVariables
- Benefits: No duplicate content bugs, no keyword leakage between levels, more efficient parsing, explicit state-driven logic
Multi-line hints and tooltips for options: Fixed option-level delimited content parsing
- parseOptions and parseBreakdownOptions now handle multi-line HINT/TOOLTIP content correctly
- Delimited content collection moved outside the - prefix check to capture all lines
- Supports option-level syntax: - HINT: """ followed by multi-line content then """
- identifyQuestions updated to skip delimited keyword blocks when identifying question boundaries
CUSTOM keyword support: Added missing parser support for custom subtotal calculations
- CUSTOM: keyword now recognized as breakdown option metadata
- Enables custom expressions for SUBTOTAL calculations: CUSTOM: {{salary + bonus}}
- Previously treated as separate option row; now correctly parsed as SUBTOTAL metadata
Consistent TOTAL styling: Removed hardcoded styling from TOTAL rows
- Removed font-bold/font-semibold classes from TOTAL row rendering
- Added Markdown rendering to TOTAL labels for user-controlled styling
- TOTAL labels now support Markdown formatting like SUBTOTAL labels: **Total Revenue**
- Removed redundant hover:bg-transparent classes from all TableRow elements
Changed to more formal Dutch placeholders
- Changed "Voer je antwoord in..." to "Voer uw antwoord in...

Breaking Changes

Delimiter syntax change: Replaced triple dashes (---) with triple quotes (""") for multi-line content
- Reason: Triple dashes conflict with Markdown setext heading syntax, causing Prettier to corrupt survey files
- Old syntax: HINT: --- followed by content then ---
- New syntax: HINT: """ followed by content then """
- Impact: All multi-line HINT and TOOLTIP blocks must be updated
- Migration: Use find-and-replace to change --- on delimiter lines to """
- Benefit: Survey .md files can now be safely formatted with Prettier without corruption
Navigation syntax change: Replaced NAV: + LEVEL: with single NAVIGATION: keyword
- Old syntax: NAV: Section Name followed by LEVEL: 1
- New syntax: NAVIGATION: 1 placed after page title (page title becomes navigation label)
- Navigation is now page-based metadata instead of separate structure
- Pages have unique IDs for reliable navigation highlighting
- Single source of truth - no page duplication between BLOCKS and navigation

New Features

Navigation settings: Added setting to control whether respondents can jump to unvisited pages
- New toggle in Settings: "Allow Navigation to Unvisited Pages"
- Default behavior: respondents can only navigate to visited or current pages
- When enabled: respondents can jump to any page in the navigation sidebar
- Setting persists in localStorage across sessions
SEPARATOR support for breakdown questions: Added missing parser support for separator rows
- - SEPARATOR keyword now recognized in breakdown questions
- Creates empty rows for visual organization between option groups
- Example: Place separator between different cost categories
- Separator rows are excluded from calculations automatically
Improved computed value display in breakdown questions: Placeholder shown when values aren't available
- When CUSTOM calculations reference unavailable variables, shows – instead of €\{variable\}.000,-
- When VALUE (prefillValue) uses unavailable variables, shows – instead of escaped placeholders
- Once variables become available, calculated values display normally with proper formatting
- Applies to both SUBTOTAL rows with CUSTOM and regular options with VALUE
Interleaved text and questions: Text can now appear between questions in natural flow
- Questions end at blank lines, allowing content to be interspersed with questions
- Section structure changed to ordered items array (content and question items)
- Example:
```
Q: First question?
TEXT

Here's some explanatory text between questions.

Q: Second question?
NUMBER
```

Bug Fixes

Fixed duplicate question IDs across blocks: Question IDs are now unique across all blocks
- Bug introduced in 0.3.0 parser refactor: each block reset the question ID counter, causing questions in different blocks to have duplicate IDs (Q1, Q1, Q2 instead of Q1, Q2, Q3)
- This caused block-level SHOW_IF conditions to fail because variable derivation would pick up the wrong question's variable
- Added test suite to prevent regression
Fixed paragraph spacing in text content: Text with blank lines between paragraphs now renders correctly
- Fixed parser bug where blank lines between content paragraphs were incorrectly skipped
- Removed redundant questionLineSet check that was marking blank lines as question lines
- Added paragraph spacing CSS (mb-4) to properly space separate paragraphs
- Added test suite for paragraph spacing with 4 test cases
Restored - TEXT option modifier: This option-level modifier was lost during parser refactor
- - TEXT sets the allowsOtherText flag on options
- Allows respondents to provide free text input when selecting specific options
- Works with both multiple choice and checkbox questions
- Added test suite for option text inputs

Improvements

Markdown-based text formatting: Removed whitespace-pre-wrap CSS and let Markdown handle paragraph breaks naturally
- Removed whitespace-pre-wrap from all text rendering components (section text, question hints/tooltips, page headers)
- Added global CSS rules for paragraph and list spacing (mb-3 with last:mb-0)
- Content now uses standard Markdown convention: double newlines create separate paragraphs with proper spacing
- Benefits: cleaner code, consistent spacing, proper semantic HTML, no CSS overrides needed
- Removed redundant wrapper divs that only held the CSS class
Parser maintainability: Extracted helper functions to reduce duplication
- Added parseComputedVariables() helper - eliminated duplication between parsePage() and parseBlock()
- Added createOption() helper - reduced boilerplate in option creation across 4 locations
- Reduced parser by 28+ lines while maintaining readability
Type system cleanup: Removed redundant and confusing types
- Removed VisiblePageContent wrapper type - now use Section[] directly
- Removed MatrixOption type - matrix questions now use standard Option type
- Removed unused Subquestion fields: subtract, subtotalLabel, value
- Renamed ComputedVariables → ComputedValues for clarity (map of values, not array of definitions)
- Removed ParsedQuestion type that duplicated the Question discriminated union
Simplified parser data structures: Removed unnecessary object wrappers and dead code
- Eliminated Chunk wrapper type - chunks are now just string[] instead of { lines: string[] }
- Removed Line wrapper type - lines are now plain strings instead of { line: string, index: number }
- Removed unused index, startIndex, and endIndex fields that were assigned but never read
- Eliminated shouldParse flag that was tracked throughout the entire parsing pipeline
- Reduced parser complexity by ~60 lines of unnecessary object creation and field tracking
- Code fences in section content are preserved naturally for markdown rendering
- No functional changes - parser works exactly the same way with simpler data structures
Section titles now display: Section headings (after ##) are now rendered as visible h2 elements
- Section content field is now optional (no empty strings stored)
- Updated documentation examples to demonstrate section usage
- Fixed section content rendering bug where page titles appeared as section content

Documentation

Improved documentation page design and readability:
- Simplified Overview example to showcase markdown-like text format without BLOCKS or NAVIGATION
- Unified navbar component across all pages with proper width matching
- Improved typography: larger font sizes for primary content, muted colors reserved for secondary content only
- Consistent Usage section styling: all sections now use bullet points instead of mixed patterns
- Added visual styling for inline code elements (monospace, background, padding, rounded corners)
- Removed "Documentation" header from sidebar for cleaner design
- Fixed Overview item styling in sidebar to match other items
- Updated Questions section description to be more concrete
- Added comprehensive documentation system guidelines to CLAUDE.md
Improved documentation UX:
- Reversed example order: code appears first, then rendered result (follows standard technical documentation pattern)
- Fixed auto-scroll bug: documentation examples no longer cause page to scroll to top when navigating between survey pages
- Removed outdated text-format-guide component reference from CLAUDE.md
Added 7 standard example files demonstrating text format features:
- basic-survey.md - Common question types
- conditional-logic.md - SHOW_IF demonstrations
- matrix-questions.md - Matrix question variations
- multi-page.md - Multi-page survey with sections
- breakdown-budget.md - Basic breakdown with computed variables
- breakdown-advanced.md - Advanced breakdown features
- range-syntax.md - RANGE syntax examples

Version 0.2.2

Released December 2025

New Features

Added SHOW_IF support for matrix subquestions - rows can now be conditionally shown/hidden based on responses
Added RANGE syntax for generating numeric options (e.g., RANGE: 1-10 creates options 1, 2, 3, ..., 10)
Redesigned documentation navigation with survey-style card layout
Added dedicated /survey route for better browser back button behavior

Improvements

Browser back button now properly returns to home page from surveys
Reduced shadow on navigation sidebars from shadow-sm to shadow-xs for more subtle styling
- Navigation shadows now match input element styling for consistency

Breaking Changes

Subquestion VARIABLE syntax changed: Use - VARIABLE: (with dash) instead of VARIABLE: (without dash)
- This makes all subquestion modifiers consistent with dash prefix: - HINT:, - TOOLTIP:, - VARIABLE:, - SHOW_IF:

Bug Fixes

Fixed matrix question parser bug where questions with subquestions were incorrectly classified as multiple_choice instead of matrix type

Testing

Added Vitest testing framework
Added comprehensive parser test suite with 13 tests covering all question types
Tests validate correct question type detection and conditional subquestion behavior

Version 0.2.1

Released December 2025

Bug Fixes

Fixed React key uniqueness warnings by using array indices instead of option values as keys across all question types
Fixed matrix question parsing bug where questions with subquestions were incorrectly classified as multiple_choice instead of matrix

Improvements

Removed hardcoded bold styling from SUBTOTAL rows - users can now control formatting via Markdown
Added comprehensive BREAKDOWN documentation covering all features

Removed

Removed TOTAL_COLUMN feature - use EXCLUDE on individual rows instead for better flexibility

Version 0.2.0

Released December 2025

New Features

BREAKDOWN Question Type

Added a new question type for collecting numeric data in table format with automatic totals and subtotals.

Automatic total calculation across all options
Support for subtotal rows with automatic or custom calculations
Multi-column layouts with column-specific totals
Header rows and separator rows for visual organization
Question and option-level PREFIX and SUFFIX to add units (e.g., currency information)
Read-only calculated values using VALUE keyword
VARIABLE support for storing individual row values (including totals)
EXCLUDE keyword to display options without including in totals
SUBTRACT keyword for deductions in calculations

TOOLTIP Support

Added tooltips for contextual help.

Tooltips work at page-, section-, question-, and option-level tooltips on page titles
Multi-line tooltip support with delimiter syntax (TOOLTIP: ---)

Documentation

Added a dedicated documentation page.

Organized by topic: Survey Structure, Question Types, Dynamic Features, Customization
Live interactive previews for every feature
Side-by-side code and rendered output
Sidebar with nested navigation

Text Editor with Auto-Save

Added a text editor to draft surveys directly in the app with automatic localStorage persistence.

Improvements

Improved multi-line HINT text

HINT text can now reliably span multiple lines using triple dashes (---) as a delimiter

Navigation

Navigation visibility now defaults to on with "Hide Navigation" toggle
Improved parent navigation items - now clickable to jump to first child page
More stable hover and active states

UI/UX

Improved section spacing for better visual hierarchy

Settings

Default language changed to Dutch (nl) to match typical survey content

Developer Experience

Implemented discriminated union types for better type safety

Removed

Removed navigation validation that prevented surveys without explicit NAV declarations

Version 0.1.0

Initial Release

Core Features

Text-Based Survey Creation

Create surveys using a simple text format that gets parsed into an interactive questionnaire.

Question Types

Multiple choice (radio buttons)
Checkbox (multi-select)
Text input (single line)
Essay (multi-line textarea)
Number input
Matrix questions (table layout with rows and columns)

Survey Structure

Multi-page surveys with navigation
Sections for organizing questions
Blocks for grouping pages

Dynamic Features

Variables for storing and referencing responses
Conditional logic (SHOW_IF) for dynamic surveys
Conditional text for dynamic question wording
Computed variables with arithmetic expressions
STARTS_WITH operator for testing multiple variables

Customization

Hints for additional question context
Markdown formatting support
List formatting (bullet lists and inline lists)

File Handling

Upload .txt or .md files
Drag and drop support

Help

Load Sample Survey
Text Syntax Guide

Settings

Language selection (Dutch/English)
Navigation visibility toggle
Navigation position (left/right)