Is your feature request related to a problem? Please describe.
Anaconda contains packages for data scientist. I propose to package find-similar package there.
nltk, skipy, pymorphy3, ... have been already in anaconda

Describe the solution you'd like

• find a way to load package to the anaconda:
• • https://enterprise-docs.anaconda.com/en/latest/data-science-workflows/packages/upload.html
• • https://docs.conda.io/projects/conda-build/en/stable/install-conda-build.html
• • https://enterprise-docs.anaconda.com/en/latest/data-science-workflows/packages/build.html
• • others
• load package
• find a way to load package automatically

If we sent dictionary to find_similar function it is not working.
We need first make prepare_dictionary and wen send the result to find_similar.
The better way is to check dictionary type first and if it's dictionary call prepare dictionary silently

• run make coverage and see the report
• add tests to all blocks
• add 100% minimum coverage to GitHub action run-tests.yml

Now

Console scripts are being used.

Task

Design and develop something with user-friendly interface. Web interface for example.

To do what?

• analyze algorithm proximity
• load test data
• analyze separate parts of algorithm (tokenise for example)

There is pylint $(git ls-files '*.py') command in GitHub action file lint.yml. This command takes files in git and run pylint for them.
Then I tried to add this command to Makefile It wasn't worked.
I think where is a way to add this command to Makefile

May be user want to search with stop words like I, we, and others.
Now we always remove stop words.
Add option remove of not (default=remove I think)

• Add coverage report using pytest-cov or something else
• Save report in html format
• Save report as artifact on GitHub

Add one or two examples with different languages

There are too many arguments in core functions due to pylint report. Need to make pylint test passed

For people who like terminal we can duplicate web interface commands and terminal commands using django management commands

• Put analytic logic separately
• use it in web interface
• duplicate it in management commands

Hard code is used in many parts of the project.
We need to change some interfaces to work with any dictionaries, languages, ...

For example: https://github.com/findsimilar/find-similar/blob/main/find_similar/tokenize.py#L14 - this usefulness words work only with concrete data.

Some strange results:
"люблю есть" / "есть люблю" - 0%
"люблю еду" / "еду люблю" - 100%
"люблю спать - 0, люблю дрова - 100%
"да е" / "е да" - 0%
"д е" и "е д" - 100%

Use weight system in algorithm.
This system will not be used by default.

In this discussion: #21 (comment) we decided to use pymorphy3 instead of pymorphy2.

• do changes which was described here: #21 (reply in thread)
• run tests and coverage on python 3.11 (make coverage)
• Add python 3.11 to GitHub action. This is the action example: https://github.com/quillcraftsman/lavacactus/blob/master/.github/workflows/testing.yml#L9

• compare two
• example frequency analysis
• tokenize one

• To demonstrate main features
• To load some simple data

• Add examples folder to the project and to the pypi package.
• Add open source examples to this folder. For example films, music, medicines, ...
• Choose examples format (yml, json, ...).
• Add parameters to find_similar search to the examples.
• Check that is really open source data (not secret and open to use).

• Add pylint to GitHub actions.
• Add yaml files linter to GitHub actions.

Now we have count parameter in find_similar. It's display results count. I propose to add another parameter display_zero_results (may be better name exists).

• If this parameter is True then we will display all results
• If this parameter is False then we will not display results with cos = 0.0

• Migrate documentation on https://findsimilar.craftsman.lol
• Update Readme

Parts of speech that should be deleted during splitting a string to tokens are hardcoded. Verbs are removed by default.
It could be a problem in some cases where verbs are significant.

It may be any texts (max count 50).
The main point is different results with remove_stopwords and without it.
Put this example in example folder in yaml format.

Is your feature request related to a problem? Please describe.
In the django package django-find-similar I use adapter from django model to TextToken in find-similar.
__init__ method works well. But I need to add new behavior the cos will be saved.

Describe the solution you'd like
Here we save cos directly.
text.cos = cos
create special method or setter to do in the TextToken class:

def set_cos(cos):
    self.cos = cos

Describe alternatives you've considered
we can create python @setter. But in this way we should make cos property protected and create @getter. It's not a problem, just need more attention and replacements.

Additional context

Now documentation builds locally and in the action it's deploy to GitHub pages.
Change this behavior this way:

1. add command to Makefile to build docs locally
2. add local build to gitignore
3. build docs in GitHub action

There is language parameter in the main function find_similar: https://github.com/findsimilar/find-similar/blob/main/find_similar/core.py#L11

This parameter = 'russian' by default. This is multi-languages text support.
But is't not working it's just mock. In the code it's hardcoded with Russian language.

• Find where language parameter need to be used
• Use this parameter to create stop word or something else

• bootstrap
• add staticfiles
• add base template

Is your feature request related to a problem? Please describe.
in __eq__ method we compare two TokenTexts by id. It's old way then we have base items and analogues.
For universal way we should compare TokenTexts by text

Describe the solution you'd like

• change __eq__ method in the TokenText class

Describe alternatives you've considered

• May be in future we will need some addition parameters to comparison

Additional context

make lint command works for untracked files only (red).
Update this command to work with all new files (red and green)

• "User-friendly library to find similar objects" in the repository description
• "User-friendly library to find similar objects" in the README
• "User-friendly library to find similar objects" on the website
• "Algorithm to define similarity rating between objects" in the setup.py

Add str and repr methods to TokenText class

• create folder
• add coverage 100%
• add GitHub action with coverage
• check main tests, lints and actions
• add local find_similar package to easy change

In find-similar==1.3.0 no examples package in the find_similar package.

Then we try to use example module:
from find_similar.examples.analyze import frequency_analysis
we have this error:
ModuleNotFoundError: No module named 'yaml'

Solution:
add yaml module to requirements in setup.py

C0103 ignore error names. It was ignored because of id attribute.
There is the one problem here. If remove C0103 form .pylintrc and run linter pylint $(git ls-files '*.py') it will fail on one line.
But if we disabled this line it wouldn't work.
We need to solve this problem and remove C0103 from .pylintrc

Handle FileNotFound error then we take non-existent example

Example from API:
Input:
{ "text_to_check": "я для тебя машину", "texts": [ "я для тебя машину", "ты для меня тост" ], "remove_stopwords": false }
Output:
[ { "text": "я для тебя машину", "cos": "0.50" }, { "text": "ты для меня тост", "cos": "0.00" } ]
With true:
Input:
{ "text_to_check": "я для тебя машину", "texts": [ "я для тебя машину", "ты для меня тост" ], "remove_stopwords": true }
Output:
[ { "text": "я для тебя машину", "cos": "1.00" }, { "text": "ты для меня тост", "cos": "0.00" } ]

This is a bug or feature?