markup: add --citeproc to pandoc converter

Adds the citeproc filter to the pandoc converter.

There are several PRs for it this feature already. However, I think
simply adding `--citeproc` is the cleanest way to enable this feature,
with the option to flesh it out later, e.g., in gohugoio#7529.

Some PRs and issues attempt adding more config options to Hugo which
indirectly configure pandoc, but I think simply configuring Pandoc via
Pandoc itself is simpler, as it is already possible with two YAML
blocks -- one for Hugo, and one for Pandoc:

    ---
    title: This is the Hugo YAML block
    ---
    ---
    bibliography: assets/pandoc-yaml-block-bibliography.bib
    ...
    Document content with @citation!

There are other useful options, e.g., gohugoio#4800 attempts to use `nocite`,
which works out of the box with this PR:

    ---
    title: This is the Hugo YAML block
    ---
    ---
    bibliography: assets/pandoc-yaml-block-bibliography.bib
    nocite: |
      @*
    ...
    Document content with no citations but a full bibliography:

    ## Bibliography

Other useful options are `csl: ...` and `link-citations: true`, which
set the path to a custom CSL file and create HTML links between the
references and the bibliography.

The following issues and PRs are related:

- Add support for parsing citations and Jupyter notebooks via Pandoc and/or Goldmark extension gohugoio#6101
  Bundles multiple requests, this PR tackles citation parsing.

- WIP: Bibliography with Pandoc gohugoio#4800
  Passes the frontmatter to Pandoc and still uses
  `--filter pandoc-citeproc` instead of `--citeproc`.
- Allow configuring Pandoc gohugoio#7529
  That PR is much more extensive and might eventually supersede this PR,
  but I think --bibliography and --citeproc should be independent
  options (--bibliography should be optional and citeproc can always be
  specified).
- Pandoc - allow citeproc extension to be invoked, with bibliography. gohugoio#8610
  Similar to gohugoio#7529, gohugoio#8610 adds a new config option to Hugo.
  I think passing --citeproc and letting the users decide on the
  metadata they want to pass to pandoc is better, albeit uglier.

Author: shoeffner

Diff for: docs/content/en/content-management/formats.md

@@ -47,21 +47,21 @@ Hugo passes reasonable default arguments to these external helpers by default:
 
 - `asciidoctor`: `--no-header-footer -`
 - `rst2html`: `--leave-comments --initial-header-level=2`
-- `pandoc`: `--mathjax`
+- `pandoc`: `--mathjax` and, for pandoc >= 2.11, `--citeproc`
 
 {{% warning "Performance of External Helpers" %}}
 Because additional formats are external commands, generation performance will rely heavily on the performance of the external tool you are using. As this feature is still in its infancy, feedback is welcome.
 {{% /warning %}}
 
 ### External Helper AsciiDoc
 
-[AsciiDoc](https://github.com/asciidoc/asciidoc) implementation EOLs in Jan 2020 and is no longer supported. 
-AsciiDoc development is being continued under [Asciidoctor](https://github.com/asciidoctor). The format AsciiDoc 
+[AsciiDoc](https://github.com/asciidoc/asciidoc) implementation EOLs in Jan 2020 and is no longer supported.
+AsciiDoc development is being continued under [Asciidoctor](https://github.com/asciidoctor). The format AsciiDoc
 remains of course. Please continue with the implementation Asciidoctor.
 
 ### External Helper Asciidoctor
 
-The Asciidoctor community offers a wide set of tools for the AsciiDoc format that can be installed additionally to Hugo. 
+The Asciidoctor community offers a wide set of tools for the AsciiDoc format that can be installed additionally to Hugo.
 [See the Asciidoctor docs for installation instructions](https://asciidoctor.org/docs/install-toolchain/). Make sure that also all
 optional extensions like `asciidoctor-diagram` or `asciidoctor-html5s` are installed if required.
 
@@ -109,13 +109,65 @@ Example of how to set extensions and attributes:
         my-attribute-name = "my value"
 ```
 
-In a complex Asciidoctor environment it is sometimes helpful to debug the exact call to your external helper with all 
+In a complex Asciidoctor environment it is sometimes helpful to debug the exact call to your external helper with all
 parameters. Run Hugo with `-v`. You will get an output like
 
 ```
 INFO 2019/12/22 09:08:48 Rendering book-as-pdf.adoc with C:\Ruby26-x64\bin\asciidoctor.bat using asciidoc args [--no-header-footer -r asciidoctor-html5s -b html5s -r asciidoctor-diagram --base-dir D:\prototypes\hugo_asciidoc_ddd\docs -a outdir=D:\prototypes\hugo_asciidoc_ddd\build -] ...
 ```
 
+### External Helper Pandoc
+
+[Pandoc](https://pandoc.org) is a universal document converter and can be used to convert markdown files.
+In Hugo, Pandoc can be used for LaTeX-style math (the `--mathjax` command line option is provided):
+
+```
+---
+title: Math document
+---
+
+Some inline math: $a^2 + b^2 = c^2$.
+```
+
+This will render in your HTML as:
+
+```
+<p>Some inline math: <span class="math inline">\(a^2 + b^2 = c^2\)</span></p>
+```
+You will have to [add MathJax](https://www.mathjax.org/#gettingstarted) to your template to properly render the math.
+
+For **Pandoc >= 2.11**, you can use [citations](https://pandoc.org/MANUAL.html#extension-citations).
+One way is to employ [BibTeX files](https://en.wikibooks.org/wiki/LaTeX/Bibliography_Management#BibTeX) to cite:
+
+```
+---
+title: Citation document
+---
+---
+bibliography: assets/bibliography.bib
+...
+This is a citation: @Doe2022
+```
+
+Note that Hugo will **not** pass its metadata YAML block to Pandoc; however, it will pass the **second** meta data block, denoted with `---` and `...` to Pandoc.
+Thus, all Pandoc settings should go there.
+
+You can also add all elements from a bibliography file (without citing them explicitly) using:
+
+```
+---
+title: My Publications
+---
+---
+bibliography: assets/bibliography.bib
+nocite: |
+  @*
+...
+```
+
+It is also possible to provide a custom [CSL style](https://citationstyles.org/authors/) by passing `csl: path-to-style.csl` as a Pandoc option.
+
+
 ## Learn Markdown
 
 Markdown syntax is simple enough to learn in a single sitting. The following are excellent resources to get you up and running:

Diff for: markup/pandoc/convert.go

@@ -15,12 +15,16 @@
 package pandoc
 
 import (
+	"bytes"
+	"fmt"
+	"strings"
+	"sync"
+
 	"github.com/gohugoio/hugo/common/hexec"
 	"github.com/gohugoio/hugo/htesting"
 	"github.com/gohugoio/hugo/identity"
-	"github.com/gohugoio/hugo/markup/internal"
-
 	"github.com/gohugoio/hugo/markup/converter"
+	"github.com/gohugoio/hugo/markup/internal"
 )
 
 // Provider is the package entry point.
@@ -65,6 +69,9 @@ func (c *pandocConverter) getPandocContent(src []byte, ctx converter.DocumentCon
 		return src, nil
 	}
 	args := []string{"--mathjax"}
+	if supportsCitations(c.cfg) {
+		args = append(args[:], "--citeproc")
+	}
 	return internal.ExternallyRenderContent(c.cfg, ctx, src, binaryName, args)
 }
 
@@ -77,6 +84,51 @@ func getPandocBinaryName() string {
 	return ""
 }
 
+var versionOnce sync.Once
+var pandocVersion string
+
+// getPandocVersion parses the pandoc version output
+func getPandocVersion(cfg converter.ProviderConfig) (string, error) {
+	var err error
+
+	versionOnce.Do(func() {
+		argsv := []any{"--version"}
+
+		var out bytes.Buffer
+		argsv = append(argsv, hexec.WithStdout(&out))
+
+		cmd, err := cfg.Exec.New(pandocBinary, argsv...)
+		if err != nil {
+			pandocVersion = ""
+			return
+		}
+
+		err = cmd.Run()
+		if err != nil {
+			cfg.Logger.Errorf("%s --version: %v", pandocBinary, err)
+		}
+
+		outbytes := bytes.Replace(out.Bytes(), []byte("\r"), []byte(""), -1)
+		output := strings.Split(string(outbytes), "\n")[0]
+		pandocVersion = strings.Split(output, " ")[1]
+	})
+
+	return pandocVersion, err
+}
+
+// SupportsCitations returns true for pandoc versions >= 2.11, which include citeproc
+func supportsCitations(cfg converter.ProviderConfig) bool {
+	pandocVersion, err := getPandocVersion(cfg)
+	supportsCitations := pandocVersion >= "2.11" && err == nil
+	if htesting.SupportsAll() {
+		if !supportsCitations {
+			panic(fmt.Sprintf("pandoc %s does not support citations", pandocVersion))
+		}
+		return true
+	}
+	return supportsCitations
+}
+
 // Supports returns whether Pandoc is installed on this computer.
 func Supports() bool {
 	hasBin := getPandocBinaryName() != ""

Diff for: markup/pandoc/convert_test.go

@@ -25,18 +25,124 @@ import (
 	qt "github.com/frankban/quicktest"
 )
 
-func TestConvert(t *testing.T) {
+func setupTestConverter(t *testing.T) (*qt.C, converter.Converter, converter.ProviderConfig) {
 	if !Supports() {
 		t.Skip("pandoc not installed")
 	}
 	c := qt.New(t)
 	sc := security.DefaultConfig
 	sc.Exec.Allow = security.NewWhitelist("pandoc")
-	p, err := Provider.New(converter.ProviderConfig{Exec: hexec.New(sc), Logger: loggers.NewErrorLogger()})
+	cfg := converter.ProviderConfig{Exec: hexec.New(sc), Logger: loggers.NewErrorLogger()}
+	p, err := Provider.New(cfg)
 	c.Assert(err, qt.IsNil)
 	conv, err := p.New(converter.DocumentContext{})
 	c.Assert(err, qt.IsNil)
-	b, err := conv.Convert(converter.RenderContext{Src: []byte("testContent")})
+	return c, conv, cfg
+}
+
+func TestConvert(t *testing.T) {
+	c, conv, _ := setupTestConverter(t)
+	output, err := conv.Convert(converter.RenderContext{Src: []byte("testContent")})
 	c.Assert(err, qt.IsNil)
-	c.Assert(string(b.Bytes()), qt.Equals, "<p>testContent</p>\n")
+	c.Assert(string(output.Bytes()), qt.Equals, "<p>testContent</p>\n")
+}
+
+func runCiteprocTest(t *testing.T, content string, expected string) {
+	c, conv, cfg := setupTestConverter(t)
+	if !supportsCitations(cfg) {
+		t.Skip("pandoc does not support citations")
+	}
+	output, err := conv.Convert(converter.RenderContext{Src: []byte(content)})
+	c.Assert(err, qt.IsNil)
+	c.Assert(string(output.Bytes()), qt.Equals, expected)
+}
+
+func TestGetPandocVersionCallTwice(t *testing.T) {
+	c, _, cfg := setupTestConverter(t)
+
+	version1, err1 := getPandocVersion(cfg)
+	version2, err2 := getPandocVersion(cfg)
+	c.Assert(version1, qt.Equals, version2)
+	c.Assert(err1, qt.IsNil)
+	c.Assert(err2, qt.IsNil)
+}
+
+func TestCiteprocWithHugoMeta(t *testing.T) {
+	content := `
+---
+title: Test
+published: 2022-05-30
+---
+testContent
+`
+	expected := "<p>testContent</p>\n"
+	runCiteprocTest(t, content, expected)
+}
+
+func TestCiteprocWithPandocMeta(t *testing.T) {
+	content := `
+---
+---
+---
+...
+testContent
+`
+	expected := "<p>testContent</p>\n"
+	runCiteprocTest(t, content, expected)
+}
+
+func TestCiteprocWithBibliography(t *testing.T) {
+	content := `
+---
+---
+---
+bibliography: testdata/bibliography.bib
+...
+testContent
+`
+	expected := "<p>testContent</p>\n"
+	runCiteprocTest(t, content, expected)
+}
+
+func TestCiteprocWithExplicitCitation(t *testing.T) {
+	content := `
+---
+---
+---
+bibliography: testdata/bibliography.bib
+...
+@Doe2022
+`
+	expected := `<p><span class="citation" data-cites="Doe2022">Doe and Mustermann
+(2022)</span></p>
+<div id="refs" class="references csl-bib-body hanging-indent"
+role="doc-bibliography">
+<div id="ref-Doe2022" class="csl-entry" role="doc-biblioentry">
+Doe, Jane, and Max Mustermann. 2022. <span>“A Treatise on Hugo
+Tests.”</span> <em>Hugo Websites</em>.
+</div>
+</div>
+`
+	runCiteprocTest(t, content, expected)
+}
+
+func TestCiteprocWithNocite(t *testing.T) {
+	content := `
+---
+---
+---
+bibliography: testdata/bibliography.bib
+nocite: |
+  @*
+...
+`
+	expected := `<div id="refs" class="references csl-bib-body hanging-indent"
+role="doc-bibliography">
+<div id="ref-Doe2022" class="csl-entry" role="doc-biblioentry">
+Doe, Jane, and Max Mustermann. 2022. <span>“A Treatise on Hugo
+Tests.”</span> <em>Hugo Websites</em>.
+</div>
+</div>
+`
+	runCiteprocTest(t, content, expected)
 }

Diff for: markup/pandoc/testdata/bibliography.bib

@@ -0,0 +1,6 @@
+@article{Doe2022,
+    author    = "Jane Doe and Max Mustermann",
+    title     = "A Treatise on Hugo Tests",
+    journal   = "Hugo Websites",
+    year      = "2022",
+}