metakgp / mftp Goto Github PK

Google groups suck (see this). PWA is too much work (see this).

There are broadly two steps:

• #81
• #82

Dockerise the codebase in for easy deployment.

So, here's the plan.

• User provides their kgpian.iitkgp.ac.in email.
• Receives an OTP and verifies the ownership of the email.
• Creates a password and completes user creation.
• Persist the session for 15 days (after this it will expire) and the user will have to relogin.

Security Perspective

• Rate limit based either on anonymous session or client IP.
• Associate an account only with 3 devices at max to prevent account sharing.

Something about the notice format changed

Implement mftp as a cronjob along with a continuously running service in infinite loop.

Basic idea

• mftp cronjob disable
• mftp cronjob enable
• mftp cronjob enable 30

Why this? and Why not only this?

Cronjob will be best for:

• Internal Servers.. not requiring OTP for login
• Small devices like rpis.. limited in resources

Service will be best for:

• External Servers.. requiring OTP for login
• Having enough resources

This is there as a TODO in the README.

I have understood most of the code now, and I think it will be easy to implement this now. How do you want to do this? This is what I have so far:

• Have a new collection in the DB that has a list of keywords and the email ID to fire an email to
• At every diffed notice, search the Notice text for the list of keywords and also the PDF attachment (using some PDF search library)
• If the keyword matches, then send that email also, otherwise stop.

I have a feeling that all this will take some time? Say 10 users, 4 keywords per user, that's 40 things to search for in the PDF. I am not sure if there's a bottleneck here, and if there is what it is.

@amrav What do you think?

Implement logs rotation to save disk storage :)
Refer: https://en.wikipedia.org/wiki/Log_rotation

• If the sanity check goes wrong, then send an email out to everyone informing them that the system may not be working right.
• discussed this with @amrav

Heroku is removing mLab addon and hence we will need to migrate from the same towards a better solution. Below are some references for the same.

• https://docs.mlab.com/shutdown-of-heroku-add-on/
• https://devcenter.heroku.com/changelog-items/1823

Here are some official resources that can be referred to:

• https://docs.mlab.com/how-to-migrate-sandbox-databases-to-atlas/
• https://docs.mlab.com/mlab-to-atlas/

NOTE: The issue needs to be fixed by 8 December 2020.

Is it possible to host this project, so that people may actually use it?
I was thinking of creating a web UI where people can subscribe to get emails from CDC Notice Board. This will be different from the already existing CDC Notify app in that the attachments in the notice will be attached to the email.
I would like to know if this is a feasible goal.

Self host custom configured ntfy instance for our needs.

Something has probably been changed here (the URL of the notice PDFs maybe?), and PDFs in our mail are incorrectly encoded.

Get notified of any abnormality, basically a sanity check. Will add and remove a cronjob which keeps checking on the status of mftp and logs and reports when it detects any abnormality.

• Implement mftp-doctor script
• Dockerise it
• Incorporate into script
  • mftp doctor enable
  • mftp doctor enable 10
  • mftp doctor disable

Google groups suck (see this). Need to develop a Progressive Web App with following necessary features:

• Chrome and Safari Support
• Notifications
• Attachment display
• #78
• Filtering. For example, Internship and Placement
• Searching. Based on anything, like, company name, date etc. Basically notification content.

Heroku-18 (currently used and the only version to support python 2) has reached its End-of-Life.
For newer builds to take place we need to upgrade to Heroku-22. So a migration of code to the latest version of Python is necessary.

Features available to individual hosters:

• Apply feature (w/ and w/o ntfy)
• Open companies list on start of the day
• #8
• Shortlist notifier
  • From Mail Body
  • From Attachment
  • Handle multiple instances of shortlists

Google account and google groups have many limitations, resulting in failure to use this as a solution to large scale newsletter service.

Refer following soruces:

• https://support.google.com/a/answer/166852?sjid=6144039545142207889-AP
• https://developers.google.com/gmail/api/reference/quota

• Check out https://docs.python.org/2/library/traceback.html

The duplicate mails are repeatedly sent with attachments of 0 bytes.

Currently, we have a cron job on the metakgp digital ocean server that pings the URL https://mftp.herokuapp.com/ and hence fetches the new notices. This way, though has been working, leads to sometimes Heroku idling the instance of the application and no ping is able to reach or somehow the metakgp cron job fails.

Solution

It would be better to use the Heroku scheduler add-on and modify the script to work with it and set a periodic frequency for the script (or function) to run and send new notices. This would make us totally independent of the cron job as well.

Ref:

• https://devcenter.heroku.com/articles/scheduler

Issue

• The README still mentions the use of mongolab while setting up, this needs to be updated as we don't use that anymore

MFTP has been showing any notices lately. It is probably because @amrav 's ERP credentials have expired. Someone new need to up ownership of the project and use her credentials, information can also be scraped form http://10.3.100.27/notice/ (local site).

The work is being done on the revamp branch.
This is the ToDo list in order of priority.

• Revamp Notice fetching with - better logic and less overhead!
• MongoDB integration
• Beautiful enough!? email formatting
• Sending Emails
• Sending Emails over LAN (campus network)
• MFTP as a service: Implement all of the functionalities of a service
• Fix currently known BUGs:
  • Attachment missing randomly - Testing on production required
  • Emails not getting sent - Testing on production required
  • Power-cut and resetting of static IP configuration
  • False positive for the session status - says dead but was alive
  • MongoDB DNS resolution error, fails after powercut. This only fails because it's the first thing to do with internet.
• Create a proper README.md
• Error Handling
  • Extensive logging

Hi, In mftp, inappropriate dependency versioning constraints can cause risks.

Below are the dependencies and version constraints that the project is using

backports-abc==0.4
backports.ssl-match-hostname==3.4.0.2
beautifulsoup4==4.4.1
certifi==2015.11.20.1
docopt==0.4.0
futures==3.0.3
pymongo==3.4
requests==2.8.1
singledispatch==3.4.0.3
six==1.10.0
tornado==4.3
wheel==0.24.0
python-dotenv==0.5.1

The version constraint == will introduce the risk of dependency conflicts because the scope of dependencies is too strict.
The version constraint No Upper Bound and * will introduce the risk of the missing API Error because the latest version of the dependencies may remove some APIs.

After further analysis, in this project,
The version constraint of dependency pymongo can be changed to >=3.0,<=4.1.1.

The above modification suggestions can reduce the dependency conflicts as much as possible,
and introduce the latest version as much as possible without calling Error in the projects.

The invocation of the current project includes all the following methods.

bson.json_util.loads
pymongo.MongoClient.get_default_database
pymongo.MongoClient.close
bson.json_util.dumps
pymongo.MongoClient

insert_from_file
further_defaulters.append
mc_old.get_default_database.notices.find
defaulters.append
start_database_export
pymongo.MongoClient.close
further_repeated.append
pymongo.MongoClient.get_default_database
open
mc_new.get_default_database.notices.insert
os.path.dirname
bson.json_util.dumps
pymongo.MongoClient
dotenv.load_dotenv
argparse.ArgumentParser.add_argument
bson.json_util.loads
argparse.ArgumentParser.add_mutually_exclusive_group
os.path.join
len
parser.add_mutually_exclusive_group.add_argument
f.write
argparse.ArgumentParser
argparse.ArgumentParser.parse_args
f.read
export_db
format
print
repeated_notices.append
insert_notice

@developer
Could please help me check this issue?
May I pull a request to fix it?
Thank you very much.

Some notices are sent repeatedly.

https://github.com/metakgp/mftp/blob/master/update.py#L49

The data needs to be extracted from the Text portion of the message which is inside notice['text'].The data is in HTML markup as:

<b>Type</b> : PLACEMENT or <b>Type</b> : INTERNSHIP.

@zorroblue @DefCon-007 I am sure this is easy. See if you can put it together really quickly. I will be off trying to get this out of BS4 meanwhile.

Implement algorithm to send notifications to subscribers (topic in case of ntfy or separate google groups) based on certain filters like:

• Internship
• Placement
• PPO

It would be better if the subject is changed from Notice: CV Submission - Company to Internship: CV Submission - Company or Placement: Urgent - CV Verification.

This is possible since the first line mentions type:placement or type:internship. This would help reduce confusions amongst people sitting for placements who might confuse notices of internship as being of placements.

MFTP currently fails to run, probably because of some changes to ERP's TNP noticeboard. The error log is:

Unhandled error occured : 
Traceback (most recent call last): 
  File "main.py", line 27, in func 
    update.check_notices() 
  File "/app/erp.py", line 78, in wrapped_func 
    *args, **kwargs) 
  File "/app/erp.py", line 93, in wrapped_func 
    func(session=session, sessionData=sessionData, *args, **kwargs) 
  File "/app/update.py", line 48, in check_notices 
    m = re.search(r'ViewNotice\("(.+?)","(.+?)"\)', a.attrs['onclick'])
KeyError: 'onclick'

metakgp/naarad#4

After the above is done, need to provide relevant credentials to be able to post to any channel.

As per github the following dependencies need to be updated and are currently vulnerable:

• requests

A few of the notices on the ERP notice board have a PDF file attached to them. It would be nice to have these attachments also available in the MFTP forums.

Context

If and when #38 is implemented using Heroku Scheduler, we will need a better way to get notified when the service is not working. Here are key things to note:

1. When we are running on the Heroku scheduler, there is a "very rare" chance of the job not running or getting missed.
2. The above is taken care of by Heroku itself, and we do not have to worry about the same. So, the fault we should be notified of is when the script runs but fails.

Requirement

A localized (can pinpoint the exact place that threw the error) and minimal (so that the maintainers do not mark it as spam and multiple maintainers can be added). A good way would be to use the existing mailing mechanism setup using Mailgun REST API. We should also aim at keeping the configuration and reconfiguration overhead as low as possible, if feasible, we should pass it from Github itself.

NOTE: this should ideally be done after #38

Resources

• For python exception handling: https://www.programiz.com/python-programming/exception-handling

The r.history array becomes empty and thus this line throws this error:

Traceback (most recent call last):
  File "update.py", line 132, in <module>
    check_notices()
  File "path/erp.py", line 77, in wrapped_func
    r.history[1].headers['Location']).group(1)
IndexError: list index out of range

Instead, if before accessing r.history[1], it should ensure that r.history's size is at least 2 and if the size is less than 2, it should ask the user to check their secret answer settings (probably a typo in one of the answers)