There are some breaking changes in this commit, see #11455.
Closes#11455Closes#11549
This fixes a set of bugs (see issue list) and it is also paying some technical debt accumulated over the years. We now build with Staticcheck enabled in the CI build.
The performance should be about the same as before for regular sized Hugo sites, but it should perform and scale much better to larger data sets, as objects that uses lots of memory (e.g. rendered Markdown, big JSON files read into maps with transform.Unmarshal etc.) will now get automatically garbage collected if needed. Performance on partial rebuilds when running the server in fast render mode should be the same, but the change detection should be much more accurate.
A list of the notable new features:
* A new dependency tracker that covers (almost) all of Hugo's API and is used to do fine grained partial rebuilds when running the server.
* A new and simpler tree document store which allows fast lookups and prefix-walking in all dimensions (e.g. language) concurrently.
* You can now configure an upper memory limit allowing for much larger data sets and/or running on lower specced PCs.
We have lifted the "no resources in sub folders" restriction for branch bundles (e.g. sections).
Memory Limit
* Hugos will, by default, set aside a quarter of the total system memory, but you can set this via the OS environment variable HUGO_MEMORYLIMIT (in gigabytes). This is backed by a partitioned LRU cache used throughout Hugo. A cache that gets dynamically resized in low memory situations, allowing Go's Garbage Collector to free the memory.
New Dependency Tracker: Hugo has had a rule based coarse grained approach to server rebuilds that has worked mostly pretty well, but there have been some surprises (e.g. stale content). This is now revamped with a new dependency tracker that can quickly calculate the delta given a changed resource (e.g. a content file, template, JS file etc.). This handles transitive relations, e.g. $page -> js.Build -> JS import, or $page1.Content -> render hook -> site.GetPage -> $page2.Title, or $page1.Content -> shortcode -> partial -> site.RegularPages -> $page2.Content -> shortcode ..., and should also handle changes to aggregated values (e.g. site.Lastmod) effectively.
This covers all of Hugo's API with 2 known exceptions (a list that may not be fully exhaustive):
Changes to files loaded with template func os.ReadFile may not be handled correctly. We recommend loading resources with resources.Get
Changes to Hugo objects (e.g. Page) passed in the template context to lang.Translate may not be detected correctly. We recommend having simple i18n templates without too much data context passed in other than simple types such as strings and numbers.
Note that the cachebuster configuration (when A changes then rebuild B) works well with the above, but we recommend that you revise that configuration, as it in most situations should not be needed. One example where it is still needed is with TailwindCSS and using changes to hugo_stats.json to trigger new CSS rebuilds.
Document Store: Previously, a little simplified, we split the document store (where we store pages and resources) in a tree per language. This worked pretty well, but the structure made some operations harder than they needed to be. We have now restructured it into one Radix tree for all languages. Internally the language is considered to be a dimension of that tree, and the tree can be viewed in all dimensions concurrently. This makes some operations re. language simpler (e.g. finding translations is just a slice range), but the idea is that it should also be relatively inexpensive to add more dimensions if needed (e.g. role).
Fixes#10169Fixes#10364Fixes#10482Fixes#10630Fixes#10656Fixes#10694Fixes#10918Fixes#11262Fixes#11439Fixes#11453Fixes#11457Fixes#11466Fixes#11540Fixes#11551Fixes#11556Fixes#11654Fixes#11661Fixes#11663Fixes#11664Fixes#11669Fixes#11671Fixes#11807Fixes#11808Fixes#11809Fixes#11815Fixes#11840Fixes#11853Fixes#11860Fixes#11883Fixes#11904Fixes#7388Fixes#7425Fixes#7436Fixes#7544Fixes#7882Fixes#7960Fixes#8255Fixes#8307Fixes#8863Fixes#8927Fixes#9192Fixes#9324
Removes emoji code conversion from the page and shortcode parsers. Emoji
codes in markdown are now passed to Goldmark, where the goldmark-emoji
extension converts them to decimal numeric character references.
This disables emoji rendering for the alternate content formats: html,
asciidoc, org, pandoc, and rst.
Fixes#7332Fixes#11587Closes#11598
This also speeds up situations where you only need the fragments/toc and not the rendered content, e.g. Related
with fragments type indexing:
```bash
name old time/op new time/op delta
RelatedSite-10 12.3ms ± 2% 10.7ms ± 1% -12.95% (p=0.029 n=4+4)
name old alloc/op new alloc/op delta
RelatedSite-10 38.6MB ± 0% 38.2MB ± 0% -1.08% (p=0.029 n=4+4)
name old allocs/op new allocs/op delta
RelatedSite-10 117k ± 0% 115k ± 0% -1.36% (p=0.029 n=4+4)
```
Fixes#10750
The main topic of this commit is that you can now index fragments (content heading identifiers) when calling `.Related`.
You can do this by:
* Configure one or more indices with type `fragments`
* The name of those index configurations maps to an (optional) front matter slice with fragment references. This allows you to link
page<->fragment and page<->page.
* This also will index all the fragments (heading identifiers) of the pages.
It's also possible to use type `fragments` indices in shortcode, e.g.:
```
{{ $related := site.RegularPages.Related .Page }}
```
But, and this is important, you need to include the shortcode using the `{{<` delimiter. Not doing so will create infinite loops and timeouts.
This commit also:
* Adds two new methods to Page: Fragments (can also be used to build ToC) and HeadingsFiltered (this is only used in Related Content with
index type `fragments` and `enableFilter` set to true.
* Consolidates all `.Related*` methods into one, which takes either a `Page` or an options map as its only argument.
* Add `context.Context` to all of the content related Page API. Turns out it wasn't strictly needed for this particular feature, but it will
soon become usefil, e.g. in #9339.
Closes#10711
Updates #9339
Updates #10725
We do lazy initialization and (potentially) reuse of an output format's rendered content. We do this evaluation when we
start a new rendering a new output format. There are, however, situation where these borders gets crossed (e.g.
accessing content from another output format). We have a check for this in place for most cases, but not the content
rendering of inner markdown blocks inside shortcodes. This patch applies that same logic to the newly introduced
RenderContent method (which is not available from the templates).
Fixes#10391
This issue was introduced in `v0.102.0`.
In 223bf28004 we removed the byte source from the parsed page result, which
meant we had to preserve exact positioning for all elements. This introduced some new `TypeIgnore` tokens
which we, wrongly, assumed didn't matter where we put in the result slice (they should be ignored anyway).
But it seems that this broke the logic where we determine if it's positional or named params in the case
where the paramater value contains escaped quoutes.
This commit makes sure that these ignore tokens (the back slashes) are never sent back to the client, which is how it was before `v0.102.0`.
This commit also fixes some lost error information in that same commit.
Fixes#10236
This commit adds a new `.InnerDeindent` method to the shortcode context, which is `.Inner` with any
indendation removed. This is then used in the built-in `highlight` shortcode to prevent the extra
whitespace getting hightlighted.
Fixes#4717
* Record the leading whitespace (tabs, spaces) before the shortcode when parsing the page.
* Apply that indentation to the rendered result of shortcodes without inner content (where the user will apply indentation).
Fixes#9946
It has been deprecated for a long time, its v1 version is not maintained anymore, and there are many known issues. Goldmark should be
a mature replacement by now.
Closes#9934
Note that we will now fail if `inlineImports` is enabled and we cannot resolve an import.
You can work around this by either:
* Use url imports or imports with media queries.
* Set `skipInlineImportsNotFound=true` in the options
Also get the argument order in the different NewFileError* funcs in line.
Fixes#9895
* Redo the server error template
* Always add the content file context if relevant
* Remove some now superflous error string matching
* Move the server error template to _server/error.html
* Add file context (with position) to codeblock render blocks
* Improve JS build errors
Fixes#9892Fixes#9891Fixes#9893
* Add file context to minifier errors when publishing
* Misc fixes (see issues)
* Allow custom server error template in layouts/server/error.html
To get to this, this commit also cleans up and simplifies the code surrounding errors and files. This also removes the usage of `github.com/pkg/errors`, mostly because of https://github.com/pkg/errors/issues/223 -- but also because most of this is now built-in to Go.
Fixes#9852Fixes#9857Fixes#9863
This ommmit contains some security hardening measures for the Hugo build runtime.
There are some rarely used features in Hugo that would be good to have disabled by default. One example would be the "external helpers".
For `asciidoctor` and some others we use Go's `os/exec` package to start a new process.
These are a predefined set of binary names, all loaded from `PATH` and with a predefined set of arguments. Still, if you don't use `asciidoctor` in your project, you might as well have it turned off.
You can configure your own in the new `security` configuration section, but the defaults are configured to create a minimal amount of site breakage. And if that do happen, you will get clear instructions in the loa about what to do.
The default configuration is listed below. Note that almost all of these options are regular expression _whitelists_ (a string or a slice); the value `none` will block all.
```toml
[security]
enableInlineShortcodes = false
[security.exec]
allow = ['^dart-sass-embedded$', '^go$', '^npx$', '^postcss$']
osEnv = ['(?i)^(PATH|PATHEXT|APPDATA|TMP|TEMP|TERM)$']
[security.funcs]
getenv = ['^HUGO_']
[security.http]
methods = ['(?i)GET|POST']
urls = ['.*']
```
Currently, no name shortcodes (`{{< >}}`) enter unexpected branch and
throw `BUG: template info not set`. This patch checks if shortcode has
name or not earlier and throws specific error.
Closes#6866
This change is mostly motivated to get a more stable CI build (we're building the Hugo site there, with Instagram and Twitter shortcodes sometimes failing).
Fixes#7866
This means that any HTML file inside /content will be treated as a regular file.
If you want it processes with shortcodes and a layout, add front matter.
The defintion of an HTML file here is:
* File with extension .htm or .html
* With first non-whitespace character "<" that isn't a HTML comment.
This is in line with the documentation.
Fixes#7030Fixes#7028
See #6789
This more or less completes the simplification of the template handling code in Hugo started in v0.62.
The main motivation was to fix a long lasting issue about a crash in HTML content files without front matter.
But this commit also comes with a big functional improvement.
As we now have moved the base template evaluation to the build stage we now use the same lookup rules for `baseof` as for `list` etc. type of templates.
This means that in this simple example you can have a `baseof` template for the `blog` section without having to duplicate the others:
```
layouts
├── _default
│ ├── baseof.html
│ ├── list.html
│ └── single.html
└── blog
└── baseof.html
```
Also, when simplifying code, you often get rid of some double work, as shown in the "site building" benchmarks below.
These benchmarks looks suspiciously good, but I have repeated the below with ca. the same result. Compared to master:
```
name old time/op new time/op delta
SiteNew/Bundle_with_image-16 13.1ms ± 1% 10.5ms ± 1% -19.34% (p=0.029 n=4+4)
SiteNew/Bundle_with_JSON_file-16 13.0ms ± 0% 10.7ms ± 1% -18.05% (p=0.029 n=4+4)
SiteNew/Tags_and_categories-16 46.4ms ± 2% 43.1ms ± 1% -7.15% (p=0.029 n=4+4)
SiteNew/Canonify_URLs-16 52.2ms ± 2% 47.8ms ± 1% -8.30% (p=0.029 n=4+4)
SiteNew/Deep_content_tree-16 77.9ms ± 1% 70.9ms ± 1% -9.01% (p=0.029 n=4+4)
SiteNew/Many_HTML_templates-16 43.0ms ± 0% 37.2ms ± 1% -13.54% (p=0.029 n=4+4)
SiteNew/Page_collections-16 58.2ms ± 1% 52.4ms ± 1% -9.95% (p=0.029 n=4+4)
name old alloc/op new alloc/op delta
SiteNew/Bundle_with_image-16 3.81MB ± 0% 2.22MB ± 0% -41.70% (p=0.029 n=4+4)
SiteNew/Bundle_with_JSON_file-16 3.60MB ± 0% 2.01MB ± 0% -44.20% (p=0.029 n=4+4)
SiteNew/Tags_and_categories-16 19.3MB ± 1% 14.1MB ± 0% -26.91% (p=0.029 n=4+4)
SiteNew/Canonify_URLs-16 70.7MB ± 0% 69.0MB ± 0% -2.40% (p=0.029 n=4+4)
SiteNew/Deep_content_tree-16 37.1MB ± 0% 31.2MB ± 0% -15.94% (p=0.029 n=4+4)
SiteNew/Many_HTML_templates-16 17.6MB ± 0% 10.6MB ± 0% -39.92% (p=0.029 n=4+4)
SiteNew/Page_collections-16 25.9MB ± 0% 21.2MB ± 0% -17.99% (p=0.029 n=4+4)
name old allocs/op new allocs/op delta
SiteNew/Bundle_with_image-16 52.3k ± 0% 26.1k ± 0% -50.18% (p=0.029 n=4+4)
SiteNew/Bundle_with_JSON_file-16 52.3k ± 0% 26.1k ± 0% -50.16% (p=0.029 n=4+4)
SiteNew/Tags_and_categories-16 336k ± 1% 269k ± 0% -19.90% (p=0.029 n=4+4)
SiteNew/Canonify_URLs-16 422k ± 0% 395k ± 0% -6.43% (p=0.029 n=4+4)
SiteNew/Deep_content_tree-16 401k ± 0% 313k ± 0% -21.79% (p=0.029 n=4+4)
SiteNew/Many_HTML_templates-16 247k ± 0% 143k ± 0% -42.17% (p=0.029 n=4+4)
SiteNew/Page_collections-16 282k ± 0% 207k ± 0% -26.55% (p=0.029 n=4+4)
```
Fixes#6716Fixes#6760Fixes#6768Fixes#6778
This commit also
* revises the change detection for templates used by content files in server mode.
* Adds a Page.RenderString method
Fixes#6545Fixes#4663Closes#6043
This is a big commit, but it deletes lots of code and simplifies a lot.
* Resolving the template funcs at execution time means we don't have to create template clones per site
* Having a custom map resolver means that we can remove the AST lower case transformation for the special lower case Params map
Not only is the above easier to reason about, it's also faster, especially if you have more than one language, as in the benchmark below:
```
name old time/op new time/op delta
SiteNew/Deep_content_tree-16 53.7ms ± 0% 48.1ms ± 2% -10.38% (p=0.029 n=4+4)
name old alloc/op new alloc/op delta
SiteNew/Deep_content_tree-16 41.0MB ± 0% 36.8MB ± 0% -10.26% (p=0.029 n=4+4)
name old allocs/op new allocs/op delta
SiteNew/Deep_content_tree-16 481k ± 0% 410k ± 0% -14.66% (p=0.029 n=4+4)
```
This should be even better if you also have lots of templates.
Closes#6594
This commmit prepares for the addition of Goldmark as the new Markdown renderer in Hugo.
This introduces a new `markup` package with some common interfaces and each implementation in its own package.
See #5963
This means that you now can do:
{{< vidur 9KvBeKu false true 32 3.14 >}}
And the boolean and numeric values will be converted to `bool`, `int` and `float64`.
If you want these to be strings, they must be quoted:
{{< vidur 9KvBeKu "false" "true" "32" "3.14" >}}
Fixes#6371
The main motivation of this commit is to add a `page.Page` interface to replace the very file-oriented `hugolib.Page` struct.
This is all a preparation step for issue #5074, "pages from other data sources".
But this also fixes a set of annoying limitations, especially related to custom output formats, and shortcodes.
Most notable changes:
* The inner content of shortcodes using the `{{%` as the outer-most delimiter will now be sent to the content renderer, e.g. Blackfriday.
This means that any markdown will partake in the global ToC and footnote context etc.
* The Custom Output formats are now "fully virtualized". This removes many of the current limitations.
* The taxonomy list type now has a reference to the `Page` object.
This improves the taxonomy template `.Title` situation and make common template constructs much simpler.
See #5074Fixes#5763Fixes#5758Fixes#5090Fixes#5204Fixes#4695Fixes#5607Fixes#5707Fixes#5719Fixes#3113Fixes#5706Fixes#5767Fixes#5723Fixes#5769Fixes#5770Fixes#5771Fixes#5759Fixes#5776Fixes#5777Fixes#5778
This avoids double parsing the page content when `enableEmoji=true`.
This commit also adds some general improvements to the parser, making it in general much faster:
```bash
benchmark old ns/op new ns/op delta
BenchmarkShortcodeLexer-4 90258 101730 +12.71%
BenchmarkParse-4 148940 15037 -89.90%
benchmark old allocs new allocs delta
BenchmarkShortcodeLexer-4 456 700 +53.51%
BenchmarkParse-4 28 33 +17.86%
benchmark old bytes new bytes delta
BenchmarkShortcodeLexer-4 69875 81014 +15.94%
BenchmarkParse-4 8128 8304 +2.17%
```
Running some site benchmarks with Emoji support turned on:
```bash
benchmark old ns/op new ns/op delta
BenchmarkSiteBuilding/TOML,num_langs=3,num_pages=5000,tags_per_page=5,shortcodes,render-4 924556797 818115620 -11.51%
benchmark old allocs new allocs delta
BenchmarkSiteBuilding/TOML,num_langs=3,num_pages=5000,tags_per_page=5,shortcodes,render-4 4112613 4133787 +0.51%
benchmark old bytes new bytes delta
BenchmarkSiteBuilding/TOML,num_langs=3,num_pages=5000,tags_per_page=5,shortcodes,render-4 426982864 424363832 -0.61%
```
Fixes#5534
An inline shortcode's name must end with `.inline`, all lowercase.
E.g.:
```bash
{{< time.inline >}}{{ now }}{{< /time.inline >}}
```
The above will print the current date and time.
Note that an inline shortcode's inner content is parsed and executed as a Go text template with the same context as a regular shortcode template.
This means that the current page can be accessed via `.Page.Title` etc. This also means that there are no concept of "nested inline shortcodes".
The same inline shortcode can be reused later in the same content file, with different params if needed, using the self-closing syntax:
```
{{< time.inline />}}
```
Fixes#4011
