0.6.0

This release makes filter chains preserve semantic value types. Review templates that serialize transformed values to YAML or depend on the exact text produced by collection filters.

Breaking changes

• yaml and yaml_property now quote all strings, including "null", "true", and "123". Typed nulls, booleans, and numbers remain unquoted. Templates that intentionally store JSON text can decode it with parse_json before serialization. Closes #11
• calc, round, and length now return numbers instead of numeric text. Downstream assignments and serializers preserve that type, so YAML such as price: "42.57" becomes price: 42.57. number_format continues to return text.
• Numeric parsing is stricter. Blank calc input and partially numeric round input such as "42abc" are preserved with a warning instead of being coerced. Non-finite arithmetic results preserve the original input rather than emitting Infinity or null.
• round preserves collection structure and converts successfully rounded numeric-string members to numbers.
• length counts singleton arrays as collections, so [42] | length changes from 2 to 1.
• first and last preserve the selected value's type. Selected objects and nested arrays now render as JSON, and empty arrays return null instead of [].
• nth returns a typed subset array. A singleton primitive result now unwraps during interpolation and before downstream text filters, so [1, 2] | nth:2 renders as 2 instead of [2].
• The exported applyFiltersWithRegistry helper now renders a final null result as empty text instead of "null", matching template interpolation.

0.5.1

Attach VS Code and Sublime Text syntax packages to tagged GitHub releases.

0.5.0

• Add VS Code and Sublime Text syntax packages for Knap Markdown, with shared generated grammars, template comments, and snippets.
• Add Prism, CodeMirror, and highlight.js adapters, with the website using the shared CodeMirror stream parser.

0.4.2

Add {% comment %} syntax for unrendered comments. See Docs.

Example:

{#
This template is used for reading notes.
{{ title }} is ignored inside this comment.
#}

0.4.1

CLI

• Add offline CLI help for syntax, filters, and logic tags, with per-filter and per-tag examples and help hints in template diagnostics.
• Add knap validate to check template syntax and static filter arguments without data or rendering.

0.4.0

• Add CLI with knap render and knap batch.
• Add configurable limits for template size, output size, intermediate values, work operations, and nesting.
• Add allowRegex: false to make split use literal separators and disable regex searches in replace. Native regex matching remains enabled by default.
• Resolve only own data properties when reading template variables and filter paths, skipping inherited values and getters. Assign variables without invoking prototype setters.
• Escape literal link labels, image alt text, and Markdown destination delimiters. Omit javascript:, vbscript:, and data: destinations while preserving relative links and application protocols.
• Require literal tag names in replace_tags.
• Check opened batch output files before truncating them and avoid following final-path symlinks where supported.

0.3.2

• Support dot property access after array indexing, including chained paths such as cast[0].details.name and keys that share a keyword name.
• Merge through merge using the original array value rather than its serialized form, and treat a single comma-separated argument as a list.
• Preserve regular expression escapes such as \s in filter arguments while still decoding escaped quotes and backslashes.
• Remove the extra output line before standalone conditional closing tags, preserving intentional blank lines.
• Preserve intentional blank lines between loop iterations, including separate Markdown tables.
• Skip loop iterations that render nothing so a false condition cannot leave a blank separator or a trailing empty item.
• Apply whitespace control from an if tag when an elseif or else branch is taken.

0.3.0

Adds Markdown formatting, collection transformations, and YAML frontmatter generation, with typed arguments and consistent collection chaining.

New filters

Markdown

• h1 through h6 for headings.
• bold and italic, with optional underscore delimiters; strike and highlight, including named highlight colors.
• hr with before, after, and both placement.
• code for inline code or automatic multiline code blocks, and code_block for explicit fenced blocks. Both accept an optional language.
• math for inline or automatic multiline math, and math_block for explicit blocks.
• comment, escape_md, hard_break, and indent.
• embed for embedded wikilinks.
• table_pretty for padded Markdown tables, using the same UTF-16 string-length padding as Obsidian's table editor.
• yaml_property for complete YAML properties with automatic indentation.

Text and collections

• compact removes null and blank-string values while retaining 0 and false.
• encode_uri encodes URI components.
• parse_json converts JSON text into typed values.
• sort supports ascending or descending order and nested property selection.
• truncate uses grapheme-aware character limits, with the suffix counting toward the limit.
• truncate_words limits the number of words and accepts a configurable suffix.
• where filters arrays by an exact typed property value.
• sum totals numeric array values or nested object properties.

Collection handling

• Format string values recursively in typed and serialized collections with Markdown and compatible utility filters.
• Expose original typed values through context.rawValue. Collection operations preserve singleton arrays, while existing text filters retain plain singleton output.
• Support pipelines such as split | sort | join, map | bold | join, and map:"name" | reverse.
• Add property shorthand to map, including nested paths such as map:"author.name", while retaining the existing arrow syntax.
• Preserve selected value types and use null for missing properties. Use own-property lookup for map, where, and sum paths.
• Render singleton wikilinks and embeds as plain references while keeping their typed arrays available to downstream collection and YAML filters.

Arguments and validation

• Expose evaluated filter arguments through context.rawArguments in template rendering and applyFiltersWithRegistry.
• Preserve the compatibility string inputs for custom filters. Parameter validators also receive typed arguments when available.
• Validate literal arguments during template validation and dynamic arguments after resolution. Reject invalid property-path and comparison-value types.
• Preserve typed strings containing embedded quotes and commas during validation.
• Improve quoted delimiter, apostrophe, backslash, quoted regex-pattern, and parenthesized argument-list handling.
• Support grouped expressions such as {{ (title ?? "Untitled") | h1 }}.
• Initialize Intl.Segmenter when character truncation is used, so importing Knap and using unrelated filters does not require it.

YAML formatting

• Extend yaml with block formatting for arrays and objects, and yaml:flow for compact inline collections.
• Preserve nested collections, escaped multiline strings, numeric values, booleans, and explicit null.
• Render complete frontmatter properties with yaml_property, including singleton lists after wikilink formatting.

0.2.3

• Fix parsing of nth offset expressions such as n+3.

0.2.2

Add yaml filter