We can of course skip reading the entire byte slice again and again.
This was a slip in the original implementation; functionally the same,
but is slightly faster, esp. for larger data sets with many shortcodes:
```
benchmark old ns/op new ns/op delta
BenchmarkReplaceShortcodeTokens-4 15505 14753 -4.85%
benchmark old allocs new allocs delta
BenchmarkReplaceShortcodeTokens-4 1 1 +0.00%
benchmark old bytes new bytes delta
BenchmarkReplaceShortcodeTokens-4 3072 3072 +0.00%
```
It would be helpful to know whether a shortcode was called with positional or
named parameters. This commit adds a boolean `IsNamedParams` property to the
`ShortcodeWithPage` struct.
Currently a `[]byte` copy is returned. In most cases this is the safe thing to do, but we should just modify/grow the slice as needed.
This is faster and consumes less memory:
```
benchmark old ns/op new ns/op delta
BenchmarkReplaceShortcodeTokens-4 7350 4419 -39.88%
benchmark old allocs new allocs delta
BenchmarkReplaceShortcodeTokens-4 5 1 -80.00%
benchmark old bytes new bytes delta
BenchmarkReplaceShortcodeTokens-4 4816 1152 -76.08%
```
This commit is aso a small spring cleaning of duplicated code in the different `PageConvert` methods.
Fixes#1516
Even as a copy at the end is needed, this consumes way less memory on Go 1.4.2:
```benchmark old ns/op new ns/op delta
BenchmarkParsePage 145979 139964 -4.12%
BenchmarkReplaceShortcodeTokens 633574 631946 -0.26%
BenchmarkShortcodeLexer 195842 187938 -4.04%
benchmark old allocs new allocs delta
BenchmarkParsePage 87 87 +0.00%
BenchmarkReplaceShortcodeTokens 9424 9415 -0.10%
BenchmarkShortcodeLexer 274 274 +0.00%
benchmark old bytes new bytes delta
BenchmarkParsePage 141830 141830 +0.00%
BenchmarkReplaceShortcodeTokens 35219 25385 -27.92%
BenchmarkShortcodeLexer 30178 30177 -0.00%
```
See #1148
This commit replaces the regexp driven `replaceShortcodeTokens` with a handwritten one.
It wasnt't possible to handle the p-tags case without breaking performance.
This fix actually improves in that area:
```
benchmark old ns/op new ns/op delta
BenchmarkParsePage 142738 142667 -0.05%
BenchmarkReplaceShortcodeTokens 665590 575645 -13.51%
BenchmarkShortcodeLexer 176038 181074 +2.86%
benchmark old allocs new allocs delta
BenchmarkParsePage 87 87 +0.00%
BenchmarkReplaceShortcodeTokens 9631 9424 -2.15%
BenchmarkShortcodeLexer 274 274 +0.00%
benchmark old bytes new bytes delta
BenchmarkParsePage 141830 141830 +0.00%
BenchmarkReplaceShortcodeTokens 52275 35219 -32.63%
BenchmarkShortcodeLexer 30177 30178 +0.00%
```
Fixes#1148
First step to use initialisms that golint suggests,
for example:
Line 116: func GetHtmlRenderer should be GetHTMLRenderer
as see on http://goreportcard.com/report/spf13/hugo
Thanks to @bep for the idea!
Note that command-line flags (cobra and pflag)
as well as struct fields like .BaseUrl and .Url
that are used in Go HTML templates need more work
to maintain backward-compatibility, and thus
are NOT yet dealt with in this commit.
First step in fixing #959.
The flag `HTML_SMARTYPANTS_ANGLED_QUOTES` was added to Blackfriday on Black Friday. This configures rendering of double quotes as angled left and right quotes («
»).
Typical use cases would be either or, or combined, but never in the same
document. As an example would be a person from Norway; he has a blog in both
English and Norwegian (his native tongue); he would then configure Blackfriday
to use angled quotes for the Norwegian section, but keep them as reqular
double quotes for the English.
This commit adds configuration support for this new flag, configuration that can be set in the site configuration, but overridden in page front matter.
Fixes#605
- `.Ref` and `.RelRef` take a reference (the logical filename for a
page, including extension and/or a document fragment ID) and return
a permalink (or relative permalink) to the referenced document.
- If the reference is a page name (such as `about.md`), the page
will be discovered and the permalink will be returned: `/about/`
- If the reference is a page name with a fragment (such as
`about.md#who`), the page will be discovered and used to add the
`page.UniqueID()` to the resulting fragment and permalink:
`/about/#who:deadbeef`.
- If the reference is a fragment and `.*Ref` has been called from
a `Node` or `SiteInfo`, it will be returned as is: `#who`.
- If the reference is a fragment and `.*Ref` has been called from
a `Page`, it will be returned with the page’s unique ID:
`#who:deadbeef`.
- `.*Ref` can be called from either `Node`, `SiteInfo` (e.g.,
`Node.Site`), `Page` objects, or `ShortcodeWithPage` objects in
templates.
- `.*Ref` cannot be used in content, so two shortcodes have been
created to provide the functionality to content: `ref` and `relref`.
These are intended to be used within markup, like `[Who]({{% ref
about.md#who %}})` or `<a href="{{% ref about.md#who %}}">Who</a>`.
- There are also `ref` and `relref` template functions (used to create
the shortcodes) that expect a `Page` or `Node` object and the
reference string (e.g., `{{ relref . "about.md" }}` or `{{
"about.md" | ref . }}`). It actually looks for `.*Ref` as defined on
`Node` or `Page` objects.
- Shortcode handling had to use a *differently unique* wrapper in
`createShortcodePlaceholder` because of the way that the `ref` and
`relref` are intended to be used in content.
Some newly added shortcode tests compared maps in assertions.
This failed on Travis, as iteration order isn't guaranteed for maps since Go 1.
This commit fixes that by do a sort of the keys in the shortcode String() function.
This commit contains a restructuring and partial rewrite of the shortcode handling.
Prior to this commit rendering of the page content was mingled with handling of the shortcodes. This led to several oddities.
The new flow is:
1. Shortcodes are extracted from page and replaced with placeholders.
2. Shortcodes are processed and rendered
3. Page is processed
4. The placeholders are replaced with the rendered shortcodes
The handling of summaries is also made simpler by this.
This commit also introduces some other chenges:
1. distinction between shortcodes that need further processing and those who do not:
* `{{< >}}`: Typically raw HTML. Will not be processed.
* `{{% %}}`: Will be processed by the page's markup engine (Markdown or (infuture) Asciidoctor)
The above also involves a new shortcode-parser, with lexical scanning inspired by Rob Pike's talk called "Lexical Scanning in Go",
which should be easier to understand, give better error messages and perform better.
2. If you want to exclude a shortcode from being processed (for documentation etc.), the inner part of the shorcode must be commented out, i.e. `{{%/* movie 47238zzb */%}}`. See the updated shortcode section in the documentation for further examples.
The new parser supports nested shortcodes. This isn't new, but has two related design choices worth mentioning:
* The shortcodes will be rendered individually, so If both `{{< >}}` and `{{% %}}` are used in the nested hierarchy, one will be passed through the page's markdown processor, the other not.
* To avoid potential costly overhead of always looking far ahead for a possible closing tag, this implementation looks at the template itself, and is branded as a container with inner content if it contains a reference to `.Inner`
Fixes#565Fixes#480Fixes#461
And probably some others.
If content pages are fully rendered in a list page, footnotes that use
the same reference (`[^fn]`) will have duplicated anchors. This change
builds on #526 to put the page filename (`Page.File.Name`) as part of
the anchor for a footnote.
This would fix discussion [116](http://discuss.gohugo.io/t/footnote-references-are-duplicated-on-list-pages/116).
It started with wanting to move templates in template bundles and the
rest followed. I did my best to start grouping related functions
together, but there are some that I missed. There is also the method
Urlize that seems to be a special function used in both worlds. I'll
need to revisit this method.
