The web framework for inventors

I get an error here [1]

data = pickle.dumps(sdict(current.session)) TypeError: 'NoneType' object is not callable

But all the objects and variables involved contains data, in no case is it None

current.session has data, sdict and pickle existed

emmett 2.4.3 emmett-crypto 0.2.2 emmett-rest 1.3.2

[1] https://github.com/emmett-framework/emmett/blob/8bf5c23c33a0fbdad260789baa6c28e1657bd2e6/emmett/sessions.py#L141

Hi @gi0baro, I do not know where the problem may be. I will try to put the maximum information about what is happening.

Given a model with an upload field, an image in my case.

In development mode works fine. I upload the image and I can show it on some page, but in production (nginx + gunicorn) the following happens:

• The image is uploaded well to the server
• The image fails to display on the page

I did this test:

Create a tunnel to the server to access port 8000 and start the development server from weppy.

[develop]

1. I uploaded the image and I could see it without problems

[nginx] 2) Try to view image uploaded in (1) and fail. 3) I uploaded another image but could not see it

[develop] 4) I can see the image uploaded in (3)

The download function is this:

@app.route("/download/<str:filename>")
def download(filename):
    stream_dbfile(db, filename)

If I copy the url of the image and paste it in another tab, sometimes it gives: "Internal Server Error" and other times malformed image

There is no error in the nginx log.

If you need any further proof, please let me know.

Jose

So, I think during this year a couple of minor releases (1.1 and 1.2) hopefully will take place, and they probably will focus on optimization – rewriting the router to gain performance is one of these – and small enhancements on the existing features.

The key-fact of this discussion is regarding the next major release (2.0) which I really hope would be ready for the end of the year or the start of the next. Here are some ideas I got in the last months.

Things I really want into 2.0

Drop the Python 2 support With the status of Python 3 and the ecosystem, and the official drop of the support in 2020, I can't see any good reason to keep all the compatibility code inside weppy. Definitely moving to Python 3 will really simplify the internal code and avoid slow downs caused by compatibility layers.

Rewrite and separate the ORM Some people already asked me to separate the ORM into a separate project, and I think that's a good idea. But I really want to take advantage of this to rewrite the bottom layer and stop depending from the pydal library. This will surely bring some advantages, like generally improving the performance of the ORM, and the ability to extend it with new features quickly (hstore and jsonb support on postgres, for example). Even if I am one of the core contributors of pydal, moving the weppy ORM layer into a separated project will definitely avoid me the pain of keeping really old and awful code just for keeping backward compatibility and to remove a lot of codes weppy doesn't really use or need. Surely rewriting this layer will cause some disadvantages, like some incompatibilities with weppy 1.0 or loosing the compatibility with some DBMS – even if today I don't really supports them, and they actually doesn't work really well 'cause nobody supports them in pydal – but I really feel bad when I think the difficulty of adding new features to the current layer. What I can say right now is that I want to keep support for sqlite, mysql, postgresql, mssql and mongo, but the list might be expanded depending on the workload. We'll see.

Support for sockets/channels I don't really want to switch the framework when I need them, so they will definitely be in 2.0. I don't actually know if the next major version will be totally async compatible, 'cause right now weppy abuses quite a lot of thread locals, but I'll see in the next month how to deal with that.

Things nice to have in 2.0

• Enhancements to the templates system: would be nice to make it even more flexible
• Support for isomorphism/frontend components: seems like a trend right know, and it's really interesting to me (can lead to frontend validation with forms and things like that), but I need to explore this more accurately.
• Commands to generate scaffolds: can be a good point to generate different applications scaffolds for new adopters

This is almost everything I have in my plans. But I am totally open to suggestions, feature requests, feedbacks, whatever. Feel free to add your 2 cents to the discussion.

Try these examples:

The following URLs fail on URL validation of forms: www.domain.tld domain.tld http://domain.tld http://www.domain.tld http://www.domain.tld/

This works: http://domain.tld/

After I tried this again - the working URL wasn't accepted a second time. So this check is inconsistent and weirdly at the same time when this URL validation passes, all other form errors aren't shown like all other fields (which are empty in this situation) are okay even if I force them to be filled in.

There must be some bug in the form validation.

I have a problem that is not directly from emmett but with the installation of orjson.

All the server infrastructure I use is freebsd (+ jails). The orjson installation with pip does not use binaries, it is compiled. To compile you need to install some dependencies, including maturin. In the maturin installation an error occurs (I reported it) and therefore I cannot install orjson.

I've been seeing this [1]

What I want to do is temporarily modify serializers.py [2] and go back to rapidjson. This is apparently the only place where orjson is referenced. The query is if this change is enough and nothing will break.

[1] https://github.com/emmett-framework/emmett/commit/2878fbdfd0ddedbee40ed863d78fb9c6525fc93e [2] https://github.com/emmett-framework/emmett/commit/5bbbae35c8517fd147e9b76ec4486432c347ef9e

Hi @gi0baro

Emmet: 2.0.0a5 Emmet-REST: 1.0.0a1

I am trying to run a weppy app with the new emmett and emmett_rest. After fixing the imports the app compiles and starts the server. But for any request return an error. The app is exactly as I took it working from weppy.

ERROR: Application exception: Traceback (most recent call last): File "/home/jose/emmett/venv38/lib/python3.8/site-packages/emmett/asgi/handlers.py", line 282, in dynamic_handler http = await self.router.dispatch() File "/home/jose/emmett/venv38/lib/python3.8/site-packages/emmett/routing/router.py", line 217, in dispatch http_cls, output = await route.dispatcher.dispatch(request, reqargs) File "/home/jose/emmett/venv38/lib/python3.8/site-packages/emmett/routing/dispatchers.py", line 75, in dispatch await self._parallel_flow(self.flow_open) File "/home/jose/emmett/venv38/lib/python3.8/site-packages/emmett/routing/dispatchers.py", line 26, in _parallel_flow tasks = [asyncio.create_task(method()) for method in flow] File "/home/jose/emmett/venv38/lib/python3.8/site-packages/emmett/routing/dispatchers.py", line 26, in tasks = [asyncio.create_task(method()) for method in flow] File "/usr/lib/python3.8/asyncio/tasks.py", line 382, in create_task task = loop.create_task(coro) File "uvloop/loop.pyx", line 1396, in uvloop.loop.Loop.create_task TypeError: a coroutine was expected, got None

What should I consider to do the w2e migration?

Hi @gi0baro

[emmett: 2.4.9]

Give the following model for User:

class User(AuthUser):
    tablename = 'auth_users'
    belongs_to({'institucion': 'Institucion'})

    id = Field.string(default=lambda: uuid4().hex)
    avatar = Field.upload(autodelete=True)
    cuit = Field.string(length=11)
      
    validation = {
            'email': {'is': 'email', 'message': 'Ingrese el Correo Electrónico'},
            'last_name': {'presence': True, 'message': 'Ingrese el Apellido'},
            'first_name': {'presence': True, 'message': 'Ingrese el Nombre'},
            'password': {'len': {'gte': 8, 'lte': 50},
                     'message': 'La contraseña debe contener entre 8 y 50 caracteres'
                     }
    }

    fields_rw = {
            'institucion': True,
            'cuit': True
            }

    rest_rw = {
        'institucion': True,
        'registration_key': True,
        'password': (False, True)
    }

And the form to create and update

@user.route("/form")
async def form():
    
    _id = request.query_params.id

    exclude_fields = []
    if _id:
        exclude_fields = ['password']

    form = await User.form(record_id=_id,  exclude_fields=exclude_fields)
    if form.accepted:
        if 'id' in form.params:
            _id = form.params.id
              
        redirect(url('.bar', params=dict(id=_id)))

    return dict(form=form)

From the form I can create a user useless, but when I edit it, the changes are not registered. It does not give error of any kind, neither in the console nor in the browser

In this way I have working on Emmett 2.3.1 (and Weppy). I can't try this in version 2.3.1 for the use of PK as UUID, but in another old project it works.

Hi, @gi0baro !

I'm cannot find nothing about regular expressions in templates, and when i create template with utf-8 encoding and write russian text, i cannot see a valid text on template rendering. Does weppy framework supports these features?

Sorry, if i ask stupid questions, and my english.

Thank you.

@gi0baro, Based on bloggy, I added signature.py. (Attached)

Modifications to bloggy.py:

from signature import SignPipe, SignInjector

app.injectors = [SignInjector()]

@app.route("/post/<int:pid>", pipeline=[SignPipe()])
def one(pid):
   ...

@app.route("/new")
@requires(lambda: auth.has_membership('admin'), url('index'))
def new_post():
    form = Post.form()
    if form.accepted:
        redirect(url('one', form.params.id, sign=SignInjector.sign))
    return dict(form=form)

index.html

{{extend 'layout.html'}}

<a href="{{=url('new_post')}}">Create a new post</a>
<ul class="posts">
{{for post in posts:}}
    <li>
        <h2>{{=post.title}}</h2>
        <a href="{{=url('one', post.id, sign=sign)}}">Read more</a>
        <hr />
    </li>
{{pass}}
{{if not posts:}}
    <li><em>No posts here so far.</em></li>
{{pass}}
</ul>

As you can see only the one function requires that it be signed.

It happens that after entering a signed link ["Read more"] other links are signed, for example new_post. It is as if the _signature parameter appears in each url, so when the signature is created, the url already contains another signature and it is obvious that validation will always fail.

I found a solution modifying weppy / espose / url by removing the _signature parameter if it exists in params like this:

# add signature
    if '_signature' in params:
        params.pop('_signature')

    if sign:
        params['_signature'] = sign(
            path, args, params, anchor, scheme, host, language)

Can you apply this change to the repository?

bloggy.zip

Here is my simple app server,use two route /api/v1/book/isbn/<str:isbn> and /api/v1/book/<int:bookid>

when I visit http://127.0.0.1:8001/api/v1/book/isbn/1231 It works 50% work at #return 1 ，50% 404， I dont know why，It has run inside the book function already。

This is my code：

#coding=utf-8
from weppy import request
from weppy import App
from weppy.tools import service

app = App(__name__)

import random

@app.route('/api/v1/book/isbn/<str:isbn>')
@app.route('/api/v1/book/<int:bookid>')
@service.json
def book(isbn=None,bookid=None):

    if random.random()>0.5:
    #return 1
        return isbn
    #return 2
    if isbn != None:
        data = {'isbn':isbn}
    if bookid != None:
        bookid = int(bookid)
        data = {'id':bookid}

    return data

if __name__ == '__main__':
    app.debug = True
    app.run(host='0.0.0.0', port=8001)

When I set form_profile_rw for field username to (True, False) in my User(AuthUserBasic) model the username field isn't displayed. It looks like read-access is broken and only writing works because if I set it to (True, True) or (False, True) I can see a textfield to type in.

Expected behaviour is explained here: http://weppy.org/docs/0.5/dal#the-models-setup-helper

Taking into account the UUID type PK support. How is it resolved in auth_permission record_id (integer) field?

Sample of docs:

maintenance = db.Preference(name='maintenance').first()
auth.add_permission(admins, 'write', 'Preference', maintenance.id)
auth.has_permission('write', 'Preference', maintenance.id)

If Preference.id is uuid, this would not work.

Jose

I have this validation

'f1': {'in': {'dbset': lambda db: db.where(db.A.f2 == session.f2)}}

In another way:

def val_f2(db):
    print(session.f2)
    return db.where(db.A.f2 == session.f2)
    
'f1': {'in': {'dbset': lambda db: val_f2(db)}}

The problem I have is that lambda runs only once and after that first time does not come to run, then it is always calling db.where from the first condition. I can check it with the print() that only runs once

Jose

Aim: make it easy to generate forms containing also related models' fields

See https://github.com/emmett-framework/emmett/discussions/324 and https://github.com/emmett-framework/emmett/discussions/350

Some might be:

• application composition
• re-utilizable ORM models
• ORM migrations' namespaces
• advanced control of AppModule lifecycle
• routes abstract classes to facilitate code recycling/extension

Hey @gi0baro ,

I wanted to know use Emmett ORM separately from the App.

Ideally, I would use a Notebook where I would declare the DB and the Models, attach the Models to the DB, and test different queries.

Is it possible ?

Thank you! Pierre