Show and Tell - Server Side Caching

hikmat · February 23, 2021, 11:05pm

Hi Emil, this feature is awesome - thank you so much for creating it! I have a small question - why is the drawback of this approach: “you cannot use client side callbacks as they will only be able to access the key”.

Emil · February 24, 2021, 4:59am

Since the data is never transferred to the client, it cannot be used in a client side callback

GiveMeCoffee · March 11, 2021, 3:59pm

Hi there

I’m interested in using your package as I need some server side caching of data and serverside rendering too. Do you have any plans of creating a version that can be installed with Anaconda? Thanks!

Emil · March 11, 2021, 5:51pm

I don’t use conda myself, so i didn’t consider a conda release. Is there a reason why you can’t use pip?

anish · March 26, 2021, 4:00pm

How do you set the threshold when using DashProxy'? It doesn't like output_defaults. I'm using the transforms, when I use dash_extensions.enrich.Dashit complains about having multiple values for keyword argumenttransforms. I have many ServersideOutput`s so I prefer to define it globally instead of inside each callback.

Emil · March 26, 2021, 6:38pm

If you use the DashProxy object, you can pass the keywords to the constructor of the transform itself,

app = DashProxy(transforms=[
    ServersideOutputTransform(backend=mybackend),  # enable use of ServersideOutput objects
])

ljam · April 14, 2021, 2:21pm

Hi @Emil !

Many thanks for your contributions. Im finding this really helpful.
Do you know how I could use a RedisCache for caching with dash-extensions? I am using the latest version of dash-extensions. Or should I go to dash-extensions==0.0.23 which allows any flask caching backend?

Thanks,

@ljam

Emil · April 14, 2021, 2:37pm

Thanks! I would recommend using the newest version. It should be realatively easy to implement a Redis binding. Here is the file system implementation,

github.com

thedirtyfew/dash-extensions/blob/ceb625964de7b2efb583a4b638846d50c6f3dc2b/dash_extensions/enrich.py#L572



    def set(self, key, value):
        raise NotImplementedError()

    def has(self, key):
        raise NotImplementedError()


# Place store implementations here.

class FileSystemStore(FileSystemCache):

    def __init__(self, cache_dir="file_system_store", **kwargs):
        super().__init__(cache_dir, **kwargs)

    def get(self, key, ignore_expired=False):
        if not ignore_expired:
            return super().get(key)
        # TODO: This part must be implemented for each type of cache.
        filename = self._get_filename(key)
        try:

ljam · May 10, 2021, 7:07am

Hi @Emil ,

Thanks for the tip! Ill try implementing this and is it works well, i might consider adding it in a PR request

Emil · May 10, 2021, 7:54am

I added a simple Redis binding in the latest release. You can try startnummer with that one

Alex101 · June 9, 2021, 10:45am

Hey, Love this concept, Ive gone to test it out and I get a problem due to my main callback also needing a daq.BooleanSwitch however this cant be pickled. Any advice on how to get around this?

Emil · June 9, 2021, 11:04am

Can you post a MWE demonstrating the issue? Why can’t it be pickled?

Alex101 · June 9, 2021, 12:45pm

Update: I found the issue, it was that the callback I use to cache data also output the required layout (which contained the BooleanSwitch Method) after splitting out the data caching and layout selections this now works

Strange I cant seem to reproduce it in a MWE, but the actual error is “_pickle.PicklingError: Can’t pickle <class ‘BooleanSwitch’>: attribute lookup BooleanSwitch on builtins failed” if that helps?

Alex101 · July 6, 2021, 8:53am

Hi Emil,

Another issue I’m running into with this is using the @cc.cached_callback where I’m outputting data to dcc.Store but also trying to get the current State of that id. see MWE below, when I try to execute this the error I get is TypeError: object supporting the buffer API required.

Any idea on how I can get around this?

@cc.cached_callback(
    [
        Output('df_history', 'data'),
    ],
    [
        Input('url', 'pathname')
    ],
    [
       State('df_history', 'data'),
    ])
def store_data(pathname, df_history):
    if df_history is None:
        <make dataframe here>
        return dataframe 
    else:
        raise PreventUpdate

Emil · July 6, 2021, 9:18am

It seems like you are using an old version? The cached_callback syntax was deprecated some time ago.

Alex101 · July 6, 2021, 9:30am

Thanks for the info, in your original post you were using the cached_callback, would you be able to share any documentation on how this server side caching should now be implemented?

Emil · July 6, 2021, 10:01am

Up-to-date examples are available in the examples folder on github, e.g.

github.com

thedirtyfew/dash-extensions/blob/master/examples/callback_serverside_output.py

import time
import dash_core_components as dcc
import dash_html_components as html
import plotly.express as px
from dash_extensions.enrich import Dash, Output, Input, State, ServersideOutput

app = Dash(prevent_initial_callbacks=True)
app.layout = html.Div([
    html.Button("Query data", id="btn"), dcc.Dropdown(id="dd"), dcc.Graph(id="graph"),
    dcc.Loading(dcc.Store(id='store'), fullscreen=True, type="dot")
])


@app.callback(ServersideOutput("store", "data"), Input("btn", "n_clicks"))
def query_data(n_clicks):
    time.sleep(1)
    return px.data.gapminder()  # no JSON serialization here


@app.callback(Input("store", "data"), Output("dd", "options"))

This file has been truncated. show original

phuonghao145 · August 19, 2021, 4:28am

Hi @Emil ,
I’m beginner to plotting. I’m trying to integrate dash with Flask and I have this error.
Could you help me?
dash : ‘1.21.0’
dash-extensions==0.0.23

phuonghao145 · August 29, 2021, 2:50pm

Hi Emil,
I have tried ServersideOutput. It works fine. But sometimes, some of my charts get this error: cannot import name ‘ValidatorCache’ from partially initialized module ‘plotly.validator_cache’. Do you know what that is?

Emil · August 29, 2021, 3:02pm

I don’t think that is related to the Serverside output.