Awesome
FastAPI-HTMX
Extension for FastAPI to make HTMX easier to use.
FastAPI-HTMX is an opinionated extension for FastAPI to speed up development of lightly interactive web applications. FastAPI-HTMX is implemented as a decorator, so it can be used on endpoints selectively. Furthermore it reduces boilerplate for Jinja2 template handling and allows for rapid prototyping by providing convenient helpers.
Install
install via pip
:
$ pip install fastapi-htmx
install via poetry
:
$ poetry add fastapi-htmx
Usage
Getting Started
Basic example using FastAPI with fastapi-htmx
my_app/api.py
:
from pathlib import Path
from fastapi import FastAPI, Request
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx, htmx_init
app = FastAPI()
htmx_init(templates=Jinja2Templates(directory=Path("my_app") / "templates"))
@app.get("/", response_class=HTMLResponse)
@htmx("index", "index")
async def root_page(request: Request):
return {"greeting": "Hello World"}
@app.get("/customers", response_class=HTMLResponse)
@htmx("customers")
async def get_customers(request: Request):
return {"customers": ["John Doe", "Jane Doe"]}
Note that:
htmx()
got parameters, specifying the Jinja2 template to usehtmx_init()
is needed for FastAPI-HTMX to find the templates- There is no direct handling of the template needed, it only needs to be specified and the needed variables need to be returned. This way endpoints can be designed in a familiar way to standard REST endpoints in FastAPI.
- This simplifies modularizing the app later (see below) and also providing a REST API if needed. See the "Usage" section for further examples.
get_customers
does not respond with the whole web page, but only with a part of it. See the HTMX documentation on how HTMX merges partials into the current web page.
request: Request
although not used in the endpoint directly, it is currently required for the decorator to work!
The Jinja2 templates to go along with the above code need to be placed like specified in htmx_init
in my_app/templates/
in order for the example to work.
The root page my_app/templates/index.jinja2
:
<!DOCTYPE html>
<html>
<head>
<title>Hello FastAPI-HTMX</title>
</head>
<body>
<h1>{{ greeting }}</h1>
<button
hx-get="/customers"
hx-swap="innerHTML"
hx-target="#customers_list"
>
Load Data
</button>
<div id="customers_list"></div>
<script src="https://unpkg.com/htmx.org@1.9.6"></script>
</body>
</html>
The partial template to load with HTMX my_app/templates/customers.jinja2
:
<ul>
{% for customer in customers %}
<li>{{ customer }}</li>
{% endfor %}
</ul>
Main Concept
The decorator htmx
provides the following helpers:
partial_template_name
The partial template to usefull_template_name
The full page template to use when URL rewriting + history is used*_template_constructor
For DRY code, in case the logic to gather all needed variables is needed multiple times
Seeing these arguments one might ask themselves: Why all these parameters? The answer is an opinionated take on how to design modular endpoints wit partials and url-rewriting support:
The idea behind FastAPI-HTMX is to maintain a modular structure in the app and with the endpoints. Similar to a REST API with a SPA. This way the frontend can be modular as well. This majorly helps with supporting URL rewriting and the history in the frontend:
- A simple endpoint just answers with the partial.
- Without it, if the URL is rewritten and a user navigates back, reloads the page or copies the URL and opens it in another tab or shares the URL, only the partial would be shown in the browser.
To enable SPA like functionality FastAPI-HTMX uses the concept of partials and fullpages as arguments for the decorator and requires to return a dict of the needed variables. Note that returning anything else than a Mapping
like a dict in the route, leads FastAPI-HTMX to return that instead of a template.
In order to support this SPA like functionality in an app, see the following example:
my_app/api_with_constructors.py
:
from pathlib import Path
from fastapi import FastAPI, Request
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx, htmx_init
app = FastAPI()
htmx_init(templates=Jinja2Templates(directory=Path("my_app") / "templates"))
def construct_customers():
return {"customers": ["John Doe", "Jane Doe"]}
def construct_root_page():
return {
"greeting": "Hello World",
**construct_customers()
}
@app.get("/", response_class=HTMLResponse)
@htmx("index", "index")
async def root_page(request: Request):
return construct_root_page()
@app.get("/customers", response_class=HTMLResponse)
@htmx("customers", "index", construct_customers, construct_root_page)
async def get_customers(request: Request):
pass
Note that:
- The
construct_*
functions are added, they now return the dataconstruct_root_page
gathers all variables specified needed for the root page, including those for partials- This also means you must avoid naming conflicts across endpoints, so dicts can be merged.
- Costly operations can still be ignored, just use if statements in the template or similar
- The decorators arguments are extended
- The second argument is the fullpage template which is used when the endpoint is called directly (new tab, navigation or reload)
- E.g. since
construct_root_page
gathers all the data for the whole page, the whole page can be returned to the client
- E.g. since
- The other arguments are just to save some boilerplate code handling the
HX-Request
header- There is no need to use the arguments for the constructor functions, they are just for convenience. If needed the endpoint can be used for the logic as well. Especially if no URL rewriting is needed.
- The second argument is the fullpage template which is used when the endpoint is called directly (new tab, navigation or reload)
For the above code to work the my_app/templates/index.jinja2
needs to be changed as well. The changes are in the button and target div.
Changed root page my_app/templates/index.jinja2
:
<!DOCTYPE html>
<html>
<head>
<title>Hello FastAPI-HTMX</title>
</head>
<body>
<h1>{{ greeting }}</h1>
<button
hx-get="/customers"
hx-push-url="true"
hx-swap="innerHTML"
hx-target="#customers_list"
>
Load Data
</button>
<div id="customers_list">
{% include 'customers.jinja2' %}
</div>
<script src="https://unpkg.com/htmx.org@1.9.6"></script>
</body>
</html>
Note that:
hx-push-url="true"
was added to the button- The partial is now loaded by default requiring the main endpoint to also provide the needed variables like shown above
The unchanged partial my_app/templates/customers.jinja2
:
<ul>
{% for customer in customers %}
<li>{{ customer }}</li>
{% endfor %}
</ul>
To add additional partials and endpoints just repeat the same logic:
- Include the partial in the parent Jinja2 template, like the main template. A hierarchy is possible as well.
- Refactor the partials endpoints logic into a function
- Add it's return value to the parents constructor function like done above in
construct_root_page
- Add the parents template and constructor function to the partials endpoints
htmx
decorator arguments
- Add it's return value to the parents constructor function like done above in
Advanced Usage
Handling HX-Request
manually
In case the htmx()
arguments for partial and fullpage callables are not flexible enough, an endpoint can be used like usual. For a bit more convenience the HX-Request
header is easily accessible via request.hx_request
:
my_app/api_with_hx_request.py
:
from pathlib import Path
from fastapi import FastAPI
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import HXRequest, htmx, htmx_init
app = FastAPI()
htmx_init(templates=Jinja2Templates(directory=Path("my_app") / "templates"))
def my_partial(email_id: int = None):
return {"email_id": email_id}
def fullpage(email_id: int = None):
return {**my_partial(email_id)}
@app.get("/email/{email_id}", response_class=HTMLResponse)
@htmx("email_detail", "index")
def get_email(request: HXRequest, email_id: int):
if request.hx_request:
return my_partial(email_id)
else:
return fullpage(email_id)
my_app/templates/index.jinja2
:
<!DOCTYPE html>
<html><head><title>Hello FastAPI-HTMX</title></head>
<body>
<div id="email_detail">
{% include 'email_detail.jinja2' %}
</div>
<script src="https://unpkg.com/htmx.org@1.9.6"></script>
</body>
</html>
my_app/templates/email_detail.jinja2
:
<p>{{ email_id }}</p>
Filters
In order to use custom Jinja2 filters like the following, configure them like below.
my_app/templates/customer.jinja2
:
<p>{{ customer.created|datetime_format }}</p>
Add custom filters for use in Jinja2 templates:
my_app/api_template_filters.py
:
from pathlib import Path
from datetime import datetime
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx_init
def datetime_format(value: datetime, format="%H:%M %d.%m.%Y"):
return value.strftime(format) if value is not None else ""
templates = Jinja2Templates(directory=Path("my_app") / "templates")
templates.env.filters["datetime_format"] = datetime_format
htmx_init(templates=templates)
# ...
Multiple Template Directories
For bigger apps, multiple template directories might be needed. For example if the app is split into several modules connected with routers. For this case template collections can be used. With htmx_init()
multiple template collections can be defined. In each endpoint instead of the templates name (account
) the collection gets specified as well, like this: Tpl(SHOP, "account")
. This works for partials and full pages.
my_app/api_multiple_template_paths.py
:
from pathlib import Path
from fastapi import FastAPI, Request
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx, htmx_init, TemplateSpec as Tpl
SHOP = "shop"
BACKEND = "backend"
app = FastAPI()
templates = {
SHOP: Jinja2Templates(directory=Path("my_app") / SHOP / "templates"),
BACKEND: Jinja2Templates(directory=Path("my_app") / BACKEND / "templates")
}
htmx_init(templates=templates)
@app.get("/account", response_class=HTMLResponse)
@htmx(Tpl(SHOP, "account"))
async def get_account(request: Request):
pass
@app.get("/products", response_class=HTMLResponse)
@htmx(Tpl(BACKEND, "products"))
async def get_products(request: Request):
pass
my_app/shop/templates/account.jinja2
:
<h1>My Account</h1>
my_app/backend/templates/products.jinja2
:
<h1>Products</h1>
Other template file extensions for SOME endpoints
In case SOME endpoints templates got another file extension than the rest, it can be specified in the @htmx()
decorator:
my_app/api_template_file_extension.py
:
from pathlib import Path
from fastapi import FastAPI, Request
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx, htmx_init
app = FastAPI()
htmx_init(templates=Jinja2Templates(directory=Path("my_app") / "templates"))
@app.get("/customers", response_class=HTMLResponse)
@htmx("customers", template_extension="html")
async def get_customers(request: Request):
pass
# ...
my_app/templates/customers.html
:
<h1>Customer</h1>
Other template file extensions for ALL endpoints
In case ALL endpoints templates got another file extension than the default jinja2
, it can be overriden in htmx_init()
like this:
my_app/api_template_file_extension.py
:
from pathlib import Path
from fastapi import FastAPI, Request
from fastapi.responses import HTMLResponse
from fastapi.templating import Jinja2Templates
from fastapi_htmx import htmx, htmx_init
app = FastAPI()
templates = Jinja2Templates(directory=Path("my_app") / "templates")
htmx_init(templates=templates, file_extension="html")
@app.get("/customers", response_class=HTMLResponse)
@htmx("customers")
async def get_customers(request: Request):
pass
my_app/templates/customers.html
:
<h1>Customer</h1>