description: In addition to Hugo's built-in variables, you can specify your own custom data in templates or shortcodes that pull from both local and dynamic sources.
These files must be YAML, JSON, XML, or TOML files (using the `.yml`, `.yaml`, `.json`, `.xml`, or `.toml` extension). The data will be accessible as a `map` in the `.Site.Data` variable.
To access the data using the `site.Data.filename` notation, the filename must begin with an underscore or a Unicode letter, followed by zero or more underscores, Unicode letters, or Unicode digits. For example:
However, note that the theme data files are merged with the project directory taking precedence. That is, Given two files with the same name and relative path, the data in the file in the root project `data` directory will override the data from the file in the `themes/<THEME>/data` directory *for keys that are duplicated*).
Therefore, theme authors should be careful not to include data files that could be easily overwritten by a user who decides to [customize a theme][customize]. For theme-specific data items that shouldn't be overridden, it can be wise to prefix the folder structure with a namespace; e.g. `mytheme/data/<THEME>/somekey/...`. To check if any such duplicate exists, run hugo with the `-v` flag.
The keys in the map created with data templates from data files will be a dot-chained set of `path`, `filename`, and `key` in the file (if applicable).
[Jaco Pastorius](https://en.wikipedia.org/wiki/Jaco_Pastorius_discography) was a great bass player, but his solo discography is short enough to use as an example. [John Patitucci](https://en.wikipedia.org/wiki/John_Patitucci) is another bass giant.
The example below is a bit contrived, but it illustrates the flexibility of data Files. This example uses TOML as its file format with the two following data files:
*`data/jazz/bass/jacopastorius.toml`
*`data/jazz/bass/johnpatitucci.toml`
`jacopastorius.toml` contains the content below. `johnpatitucci.toml` contains a similar list:
The list of bass players can be accessed via `.Site.Data.jazz.bass`, a single bass player by adding the filename without the suffix, e.g. `.Site.Data.jazz.bass.jacopastorius`.
You can now render the list of recordings for all the bass players in a template:
For `getCSV`, the one-character-long separator must be placed in the first position followed by the URL. The following is an example of creating an HTML table in a [partial template][partials] from a published CSV:
Each downloaded URL will be cached in the default folder `$TMPDIR/hugo_cache/`. The variable `$TMPDIR` will be resolved to your system-dependent temporary directory.
With the command-line flag `--cacheDir`, you can specify any folder on your system as a caching directory.
You can also set `cacheDir` in the [main configuration file][config].
To load local files with `getJSON` and `getCSV`, the source files must reside within Hugo's working directory. The file extension does not matter, but the content does.
There is no chance to trigger a [LiveReload] when the content of a URL changes. However, when a *local* file changes (i.e., `data/*` and `themes/<THEME>/data/*`), a LiveReload will be triggered. Symlinks are not supported. Note too that because downloading data takes a while, Hugo stops processing your Markdown files until the data download has been completed.
If you change any local file and the LiveReload is triggered, Hugo will read the data-driven (URL) content from the cache. If you have disabled the cache (i.e., by running the server with `hugo server --ignoreCache`), Hugo will re-download the content every time LiveReload triggers. This can create *huge* traffic. You may reach API limits quickly.
- GitHub Starred Repositories [in a post](https://github.com/SchumacherFM/blog-cs/blob/master/content%2Fposts%2Fgithub-starred.md) using data-driven content in a [custom short code](https://github.com/SchumacherFM/blog-cs/blob/master/layouts%2Fshortcodes%2FghStarred.html).
