Skip to content

Latest commit

 

History

History
1724 lines (1276 loc) · 75.2 KB

CONTRIBUTING.md

File metadata and controls

1724 lines (1276 loc) · 75.2 KB

CONTRIBUTING

First off, thanks for taking the time to contribute (or at least read the Contributing Guidelines)! 🚀

The following is a set of guidelines for contributing to OpenBB Terminal. These are mostly guidelines, not rules. Use your best judgment, and feel free to propose changes to this document in a pull request.

BASIC

Adding a new command

Before implementing a new command we highly recommend that you go through Understand Code Structure and Follow Coding Guidelines. This will allow you to get your PR merged faster and maintain consistency in our code base.

In the next sections we describe the process to add a new command. We will be adding a function to get price targets from the Financial Modeling Prep API. Note that there already exists a function to get price targets from the Business Insider website, stocks/fa/pt, so we will be adding a new function to get price targets from the Financial Modeling Prep API, and go through adding sources.

Select Feature

  • Pick a feature you want to implement or a bug you want to fix from our issues.
  • Feel free to discuss what you'll be working on either directly on the issue or on our Discord.
    • This ensures someone from the team can help you and there isn't duplicated work.

Before writing any code, it is good to understand what the data will look like. In this case, we will be getting the price targets from the Financial Modeling Prep API, and the data will look like this:

[
  {
    "symbol": "AAPL",
    "publishedDate": "2023-02-03T16:19:00.000Z",
    "newsURL": "https://pulse2.com/apple-stock-receives-a-195-price-target-aapl/",
    "newsTitle": "Apple Stock Receives A $195 Price Target (AAPL)",
    "analystName": "Cowen Cowen",
    "priceTarget": 195,
    "adjPriceTarget": 195,
    "priceWhenPosted": 154.5,
    "newsPublisher": "Pulse 2.0",
    "newsBaseURL": "pulse2.com",
    "analystCompany": "Cowen & Co."
  }

Model

  1. Create a file with the source of data as the name followed by _model if it doesn't exist. In this case, the file openbb_terminal/stocks/fundamental_analysis/fmp_model.py already exists, so we will add the function to that file.
  2. Add the documentation header
  3. Do the necessary imports to get the data
  4. Define a function starting with get_
  5. In this function:
    1. Use type hinting
    2. Write a descriptive description where at the end the source is specified.
    3. Utilize an official API, get and return the data.
""" Financial Modeling Prep Model """
__docformat__ = "numpy"

import logging

import pandas as pd

from openbb_terminal.core.session.current_user import get_current_user
from openbb_terminal.decorators import check_api_key, log_start_end
from openbb_terminal.helpers import request

logger = logging.getLogger(__name__)

@log_start_end(log=logger)
@check_api_key(["API_KEY_FINANCIALMODELINGPREP"])
def get_price_targets(cls, symbol: str) -> pd.DataFrame:
    """Get price targets for a company [Source: Financial Modeling Prep]

    Parameters
    ----------
    symbol : str
        Symbol to get data for

    Returns
    -------
    pd.DataFrame
        DataFrame of price targets
    """
    current_user = get_current_user()

    url = f"https://financialmodelingprep.com/api/v4/price-target?symbol={symbol}&apikey={current_user.credentials.API_KEY_FINANCIALMODELINGPREP}"
    response = request(url)

    # Check if the response is valid
    if response.status_code != 200 or "Error Message" in response.json():
        message = f"Error, Status Code: {response.status_code}."
        message = (
            message
            if "Error Message" not in response.json()
            else message + "\n" + response.json()["Error Message"] + ".\n"
        )
        console.print(message)
        return pd.DataFrame()

    return pd.DataFrame(response.json())

In this function:

  • We import the current user object and, consequently, preferences using the get_current_user function. API keys are stored in current_user.credentials
  • We use the @log_start_end decorator to add the function to our logs for debugging purposes.
  • We add the @check_api_key decorator to confirm the API key is valid.
  • We have type hinting and a docstring describing the function.
  • We use the openbb_terminal helper function request, which is an abstracted version of the requests library, which allows us to add user agents, timeouts, caches, etc. to any HTTP request in the terminal.
  • We check for different error messages. This will depend on the API provider and usually requires some trial and error. With the FMP API, if there is an invalid symbol, we get a response code of 200, but the json response has an error message field. Same with an invalid API key.
  • When an error is caught, we still return an empty dataframe.
  • We return the json response as a pandas dataframe. Most functions in the terminal should return a dataframe, but if not, make sure that the return type is specified.

Note:

  1. If the function is applicable to many asset classes, it is possible that this file needs to be created under common/ directory rather than stocks/, which means the function should be written in a generic way, i.e. not mentioning stocks or a specific context.
  2. If the model requires an API key, make sure to handle the error and output relevant message.
  3. If the data provider is not yet supported, you'll most likely need to do some extra steps in order to add it to the keys menu. See this section for more details.

Some of the most common error messages are:

  • Error in the request (HTTP error)
  • Invalid API Keys
  • API Keys not authorized for Premium feature
  • Empty return payload
  • Invalid arguments (Optional)

In the example below, you can see that we explicitly handle some of them. It's not always possible to distinguish error types using status_code. So depending on the API provider, you can use either error messages or exceptions.

@log_start_end(log=logger)
@check_api_key(["API_NEWS_TOKEN"])
def get_news(
    query: str,
    limit: int = 10,
    start_date: Optional[str] = None,
    show_newest: bool = True,
    sources: str = "",
) -> pd.DataFrame:

    ...

    link += f"&apiKey={get_current_user().credentials.API_NEWS_TOKEN}"
    response = request(link)
    articles = {}

    if response.status_code == 200:
        response_json = response.json()
        articles = (response_json["articles"] if show_newest else response_json["articles"][::-1])

    elif response.status_code == 426:
        console.print(f"Error in request: {response.json()['message']}", "\n")
    elif response.status_code == 401:
        console.print("[red]Invalid API Key[/red]\n")
    elif response.status_code == 429:
        console.print("[red]Exceeded number of calls per minute[/red]\n")
    else:
        console.print(f"Error in request: {response.json()['message']}", "\n")

    ...

Click here to see the example in detail.

Data sources

Now that we have added the model function getting, we need to specify that this is an available data source. To do so, we edit the openbb_terminal/miscellaneous/sources/openbb_default.json file. This file, described below, uses a dictionary structure to identify available sources. Since we are adding FMP to stocks/fa/pt, we find that entry and append it:

    "fa": {
      "pt": ["BusinessInsider", "FinancialModelingPrep"],

If you are adding a new function with a new data source, make a new value in the file. If the data source requires an API key, please refer to the guide below for adding them. Instructions for obtaining the new api key should be included in the file OpenBBTerminal/website/content/terminal/usage/data/api-keys.md.

View

  1. Create a file with the source of data as the name followed by _view if it doesn't exist, e.g. fmp_view
  2. Add the documentation header
  3. Do the necessary imports to display the data. One of these is the _model associated with this _view. I.e. from same data source.
  4. Define a function starting with display_
  5. In this function:
    • Use typing hints
    • Write a descriptive description where at the end the source is specified
    • Get the data from the _model and parse it to be output in a more meaningful way.
    • Do not degrade the main data dataframe coming from model if there's an export flag. This is so that the export can have all the data rather than the short amount of information we may show to the user. Thus, in order to do so df_data = df.copy() can be useful as if you change df_data, df remains intact.
  6. If the source requires an API Key or some sort of tokens, add check_api_key decorator on that specific view. This will throw a warning if users forget to set their API Keys
  7. Finally, call export_data where the variables are export variable, current filename, command name, and dataframe.
@log_start_end(log=logger)
@check_api_key(["API_KEY_FINANCIALMODELINGPREP"])
def display_price_targets(
    symbol: str, limit: int = 10, export: str = "", sheet_name: Optional[str] = None
):
    """Display price targets for a given ticker. [Source: Financial Modeling Prep]

    Parameters
    ----------
    symbol : str
        Symbol
    limit: int
        Number of last days ratings to display
    export: str
        Export dataframe data to csv,json,xlsx file
    sheet_name: str
        Optionally specify the name of the sheet the data is exported to.
    """
    columns_to_show = [
        "publishedDate",
        "analystCompany",
        "adjPriceTarget",
        "priceWhenPosted",
    ]
    price_targets = fmp_model.get_price_targets(symbol)
    if price_targets.empty:
        console.print(f"[red]No price targets found for {symbol}[/red]\n")
        return
    price_targets["publishedDate"] = price_targets["publishedDate"].apply(
        lambda x: datetime.strptime(x, "%Y-%m-%dT%H:%M:%S.%fZ").strftime("%Y-%m-%d %H:%M")
    )
    export_data(
        export,
        os.path.dirname(os.path.abspath(__file__)),
        "pt",
        price_targets,
        sheet_name,
    )

    print_rich_table(
        price_targets[columns_to_show].head(limit),
        headers=["Date", "Company", "Target", "Posted Price"],
        show_index=False,
        title=f"{symbol.upper()} Price Targets",
    )

In this function:

  • We use the same log and API decorators as in the model.
  • We define the columns we want to show to the user.
  • We get the data from the fmp_model function
  • We check if there is data. If something went wrong, we don't want to show it, so we print a message and return. Note that because we have error messages in both the model and view, there will be two print outs. If you wish to just show one, it is better to handle in the model.
  • We do some parsing of the data to make it more readable. In this case, the output from FMP is not very clear at quick glance, we we put it into something more readable.
  • We export the data. In this function, I decided to export after doing the manipulation. If we do any removing of columns, we should copy the dataframe before exporting.
  • We print the data in table form using our print_rich_table. This provides a nice console print using the rich library. Note that here I show the top limit rows of the dataframe. Care should be taken to make sure that things are sorted. If a sort is required, there is a reverse argument that can be added to sort in reverse order.

Controller

Now that we have the model and views, it is time to add to the controller.

  1. Import the associated _view function to the controller.

  2. Add command name to variable CHOICES_COMMANDS from FundamentalAnalysisController class.

  3. Add command and source to print_help().

     def print_help(self):
         """Print help."""
         mt = MenuText("stocks/fa/")
         mt.add_cmd("load")
         mt.add_raw("\n")
         mt.add_param("_ticker", self.ticker.upper())
         mt.add_raw("\n")
         mt.add_info("_company_overview")
         mt.add_cmd("mktcap")
         mt.add_cmd("overview")
         mt.add_cmd("divs", not self.suffix)
    
         ...
    
         mt.add_cmd("pt")
         mt.add_cmd("dcf")
         mt.add_cmd("dcfc")
         console.print(text=mt.menu_text, menu="Stocks - Fundamental Analysis")
  4. If there is a condition to display or not the command, this is something that can be leveraged through the add_cmd method, e.g. mt.add_cmd("divs", not self.suffix).

  5. Add command description to file i18n/en.yml. Use the path and command name as key, e.g. stocks/fa/pt and the value as description. Please fill in other languages if this is something that you know.

  6. Add a method to FundamentalAnalysisController class with name: call_ followed by command name.

    • This method must start defining a parser with arguments add_help=False and formatter_class=argparse.ArgumentDefaultsHelpFormatter. In addition prog must have the same name as the command, and description should be self-explanatory ending with a mention of the data source.

    • Add parser arguments after defining parser. One important argument to add is the export capability. All commands should be able to export data.

    • If there is a single or even a main argument, a block of code must be used to insert a fake argument on the list of args provided by the user. This makes the terminal usage being faster.

      if other_args and "-" not in other_args[0][0]:
            other_args.insert(0, "-l")
    • Parse known args from list of arguments and values provided by the user.

    • Call the function contained in a _view.py file with the arguments parsed by argparse.

Note that the function self.parse_known_args_and_warn() has some additional options we can add. If the function is showing a chart, but we want the option to show raw data, we can add the raw=True keyword and the resulting namespace will have the raw attribute. Same with limit, we can pass limit=10 to add the -l flag with default=10. Here we also specify the export, and whether it is data only, plots only or anything else. This function also adds the source attribute to the namespace. In our example, this is important because we added an additional source.

Our new function will be:

   @log_start_end(log=logger)
    def call_pt(self, other_args: List[str]):
        """Process pt command"""
        parser = argparse.ArgumentParser(
            add_help=False,
            prog="pt",
            description="""Prints price target from analysts. [Source: Business Insider and Financial Modeling Prep]""",
        )
        parser.add_argument(
            "-t",
            "--ticker",
            dest="ticker",
            help="Ticker to analyze",
            type=str,
            default=None,
        )
        if other_args and "-" not in other_args[0][0]:
            other_args.insert(0, "-t")
        ns_parser = self.parse_known_args_and_warn(
            parser, other_args, EXPORT_BOTH_RAW_DATA_AND_FIGURES, raw=True, limit=10
        )
        if ns_parser:
            if ns_parser.ticker:
                self.ticker = ns_parser.ticker
                self.custom_load_wrapper([self.ticker])

            if ns_parser.source == "BusinessInsider":
                business_insider_view.display_price_target_from_analysts(
                    symbol=self.ticker,
                    data=self.stock,
                    start_date=self.start,
                    limit=ns_parser.limit,
                    raw=ns_parser.raw,
                    export=ns_parser.export,
                    sheet_name=" ".join(ns_parser.sheet_name)
                    if ns_parser.sheet_name
                    else None,
                )
            elif ns_parser.source == "FinancialModelingPrep":
                fmp_view.display_price_targets(
                    symbol=self.ticker,
                    limit=ns_parser.limit,
                    export=ns_parser.export,
                    sheet_name=" ".join(ns_parser.sheet_name)
                    if ns_parser.sheet_name
                    else None,
                )

Here, we make the parser, add the arguments, and then parse the arguments. In order to use the fact that we had a new source, we add the logic to access the correct view function. In this specific menu, we also allow the user to specify the symbol with -t, which is what the first block is doing.

Note that in the fa submenu, we allow the function to be run by specifying a ticker, ie pt -t AAPL. In this submenu we do a load behind the scenes with the ticker selected so that other functions can be run without specifying the ticker.

Now from the terminal, this function can be run as desired:

2023 Mar 03, 11:37 (🦋) /stocks/fa/ $ pt -t aapl --source FinancialModelingPrep

                         AAPL Price Targets
┏━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━┳━━━━━━━━━━━━━━┓
┃ Date             ┃ Company               ┃ Target ┃ Posted Price ┃
┡━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━╇━━━━━━━━━━━━━━┩
│ 2023-02-03 16:19 │ Cowen & Co.           │ 195.00 │ 154.50       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 09:31 │ D.A. Davidson         │ 173.00 │ 157.09       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 08:30 │ Rosenblatt Securities │ 173.00 │ 150.82       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 08:29 │ Wedbush               │ 180.00 │ 150.82       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 07:21 │ Raymond James         │ 170.00 │ 150.82       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 07:05 │ Barclays              │ 145.00 │ 150.82       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-03 03:08 │ KeyBanc               │ 177.00 │ 150.82       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-02 02:08 │ Rosenblatt Securities │ 165.00 │ 145.43       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-02 02:08 │ Deutsche Bank         │ 160.00 │ 145.43       │
├──────────────────┼───────────────────────┼────────┼──────────────┤
│ 2023-02-02 02:08 │ J.P. Morgan           │ 180.00 │ 145.43       │
└──────────────────┴───────────────────────┴────────┴──────────────┘

When adding a new menu, the code looks like this:

@log_start_end(log=logger)
def call_fa(self, _):
    """Process fa command"""
    from openbb_terminal.stocks.fundamental_analysis.fa_controller import (
        FundamentalAnalysisController,
    )

    self.queue = self.load_class(
        FundamentalAnalysisController, self.ticker, self.start, self.stock, self.queue
    )

The import only occurs inside this menu call, this is so that the loading time only happens here and not at the terminal startup. This is to avoid slow loading times for users that are not interested in stocks/fa menu.

In addition, note the self.load_class which allows to not create a new FundamentalAnalysisController instance but re-load the previous created one. Unless the arguments self.ticker, self.start or self.stock have changed since. The self.queue list of commands is passed around as it contains the commands that the terminal must perform.

Add SDK endpoint (V3)

In order to add a command to the SDK, follow these steps:

  1. If you've created a new model or view file, add the import with an alias to openbb_terminal/core/sdk/sdk_init.py following this structure:

    # Stocks - Fundamental Analysis
    from openbb_terminal.stocks.fundamental_analysis import (
        finviz_model as stocks_fa_finviz_model,
        finnhub_model as stocks_fa_finnhub_model,
        finnhub_view as stocks_fa_finnhub_view,
    )
  2. Go to the trail_map.csv file located in openbb_terminal/core/sdk, which should look like this:

    trail,model,view
    stocks.fa.analyst,stocks_fa_finviz_model.get_analyst_data,
    stocks.fa.rot,stocks_fa_finnhub_model.get_rating_over_time,stocks_fa_finnhub_view.rating_over_time
    
    ...
    
    

    In this file, the trail represents the path to the function to be called. The model represents the import alias we gave to the _model file. The view represents the import alias we gave to the _view file.

  3. Add your new function to this structure. In the below example of the pt function, our trail would be stocks.fa.pt.

    Our naming convention is such that the data source should not be included in the trail. In this example, calling a new function pt_fmp would not be allowed. For functions with multiple sources, there should be a single pt function that takes in the source as an argument. In the following example, we will stick with showing how the business_insider was initially added to the sdk.

    The model is the import alias to the _model function that was written:

    • stocks_fa_business_insider_model.get_price_target_from_analysts

    The view is the import alias to the _view function that was written:

    • stocks_fa_business_insider_view.display_price_target_from_analysts

    The added line of the file should look like this:

    stocks.fa.pt,stocks_fa_business_insider_model.get_price_target_from_analysts,stocks_fa_business_insider_view.display_price_target_from_analysts
    
  4. Generate the SDK files by running python generate_sdk.py from the root of the project. This will automatically generate the SDK openbb_terminal/sdk.py, corresponding openbb_terminal/core/sdk/controllers/ and openbb_terminal/core/sdk/models/ class files.

    To sort the trail_map.csv file and generate the SDK files, run the following command

    python generate_sdk.py sort

Add OpenBB Platform endpoint (V4)

Refer to the documentation here.

Add Unit Tests

This is a vital part of the contribution process. We have a set of unit tests that are run on every Pull Request. These tests are located in the tests folder.

Unit tests minimize errors in code and quickly find errors when they do arise. Integration tests are standard usage examples, which are also used to identify errors.

A thorough introduction on the usage of unit tests and integration tests in OpenBBTerminal can be found on the following page:

Unit Test README

Integration Test README

Any new features that do not contain unit tests will not be accepted.

Open a Pull Request

For starters, you should ensure that your branch is up to date with the develop branch. To do that, one can run the following commands:

git fetch upstream
git checkout develop
git merge upstream/develop

After that, you can create a new branch for your feature. E.g. git checkout -b feature/AmazingFeature.

Once you're happy with what you have, push your branch to remote. E.g. git push origin feature/AmazingFeature.

Note that we follow gitflow naming convention, so your branch name should be prefixed with feature/ or hotfix/ depending on the type of work you are doing. To learn more, please refer to Branch Naming Conventions.

A user may create a Draft Pull Request when there is the intention to discuss implementation with the team.

Review Process

As soon as the Pull Request is opened, our repository has a specific set of github actions that will not only run linters on the branch just pushed, but also run pytest on it. This allows for another layer of safety on the code developed.

In addition, our team is known for performing diligent code reviews. This not only allows us to reduce the amount of iterations on that code and have it to be more future proof, but also allows the developer to learn/improve his coding skills.

Often in the past the reviewers have suggested better coding practices, e.g. using 1_000_000 instead of 1000000 for better visibility, or suggesting a speed optimization improvement.

Understand Code Structure

Backend

CLI 💻 → _controller.py 🤖 → _view.py 🎨     →      _model.py 🧠
                                                      chart=True                 chart=False
                                                                    ↑    sdk.py 🏭   ↑

File                            Role Description
_controller.py 🤖 The router/input validator The controller file should hold the least amount of logic possible. Its role is to be a stupid (no logic) router and redirect the command correctly while checking the input with argparser.
_view.py 🎨 The artist The view file should only output or visualize the data it gets from the _model file! The _view can limit the data coming from the _model, otherwise the data object should be identical in the _view and the _model files.
_model.py 🧠 The brain The model file is where everything fun happens. The data is gathered (external APIs), processed and returned here.
sdk.py 🏭 The SDK Factory The SDK file is where the callable functions are created for the SDK. There is only one SDK file in the openbb_terminal folder.

Frontend

Item Description Example
CONTEXT Specific instrument world to analyse. stocks, crypto, economy
CATEGORY Group of similar COMMANDS to do on the instrument
There are specialized categories, specific to each CONTEXT and there are common categories which are not specific to one CONTEXT.
due_diligence, technical_analysis, insider
COMMAND Operation on one or no instrument that retrieves data in form of string, table or plot. rating, supplier, sentiment

The following layout is expected: /<context>/<category>/<command_files>

If there are sub-categories, the layout will be: /<context>/<category>/<sub-category>/<command_files>

Example:

openbb_terminal/stocks/stocks_controller.py
                      /stocks_helper.py
                      /technical_analysis/ta_controller.py
                                         /tradingview_view.py
                                         /tradingview_model.py
                /common/technical_analysis/overlap_view.py
                                          /overlap_model.py
                /crypto/crypto_controller.py
                       /crypto_helper.py
                       /due_diligence/dd_controller.py
                                     /binance_view.py
                                     /binance_model.py
                       /technical_analysis/ta_controller.py

With:

Context Category File Description
stocks/ stocks_controller.py Manages stocks context from a user perspective, i.e. routing commands and arguments to output data, or, more importantly, redirecting to the selected category.
stocks/ stocks_helper.py Helper to stocks menu. This file is meant to hold generic purpose stocks functionalities.
stocks/ technical_analysis/ ta_controller.py Manages technical_analysis category from stocks context from a user perspective, i.e. routing commands and arguments to output data.
stocks/ technical_analysis/ tradingview_view.py This file contains functions that rely on TradingView data. These functions represent commands that belong to technical_analysis category from stocks context. These functions are called by ta_controller.py using the arguments given by the user and will output either a string, table or plot.
stocks/ technical_analysis/ tradingview_model.py This file contains functions that rely on TradingView data. These functions represent commands that belong to technical_analysis category from stocks context. These functions are called by tradingview_view.py and will return data to be processed in either a string, dictionary or dataframe format.
common/ technical_analysis/ overlap_view.py This file contains functions that rely on overlap data. In this case overlap is not a data source, but the type of technical analysis performed. These functions represent commands that belong to technical_analysis category from MULTIPLE contexts. These functions are called by ta_controller.py, from MULTIPLE contexts, using the arguments given by the user and will output either a string, table or plot. Due to the fact that this file is common to multiple contexts the functions need to be generic enough to accommodate for this. E.g. if we are providing a dataframe to these functions, we should make sure that stocks/ta_controller.py and crypto/ta_controller use the same formatting.
common/ technical_analysis/ overlap_model.py This file contains functions that rely on overlap data. In this case overlap is not a data source, but the type of technical analysis performed. These functions represent commands that belong to technical_analysis category from MULTIPLE contexts. These functions are called by overlap_view.py, and will return data to be processed in either a string, dictionary or dataframe format. Due to the fact that this file is common to multiple contexts the functions need to be generic enough to accommodate for this. E.g. if we are getting the sentiment of an instrument, we should ensure that these functions accept both a "GME" or a "BTC", for stocks and crypto, respectively.

Follow Coding Guidelines

General Code Requirements

  1. Each function should have default values for non critical kwargs

    • Why? It increases code readability and acts as an input example for the function's arguments. This increases the ease of use of the functions through the SDK, but also just generally.

    Watch out, add default values whenever possible, but take care for not adding mutable default arguments! More info


    Good code ✅ Bad code ❌
    def display_last_uni_swaps(
      top: int = 10,
      sortby: str = "timestamp",
      descend: bool = False,
      export: str = "",) -> None:
    def display_last_uni_swaps(
      top: int,
      sortby: str,
      descend: bool,
      export: str,) -> None:

  2. Simple and understandable input objects; avoid for example weird dictionaries packed with data: {“title”: DataFrame}

    • Why? Ease of use and often these complex formats are clumsy, prone to error and the formatting required for complex parameters is time consuming and unneeded.

    Good code ✅ Bad code ❌
    def get_coins(
      top: int = 250,
      category: str = "") -> pd.DataFrame:
    def load(
      file: str,
      file_types: list,
      data_files: Dict[Any, Any],
      data_examples: Dict[Any, Any],) -> pd.DataFrame:

  3. Each function needs to have a docstring explaining what it does, its parameters and what it returns.

    • Why? You can use the function without reading its source code. This improves the developing experience and SDK usage. The SDK factory also can’t handle functions without docstrings.

  4. Consistent and clear argument naming; not symbol in _view and then ticker in _file -> ticker everywhere; the name should be descriptive of what information it holds (see Style Guide section below)

    • Why? You can quickly understand what the input should be; example: tickers and stock names are fundamentally different, but they’re both strings so they should be named accordingly.

    Good code ✅ Bad code ❌
    data: pd.Series, dataset_name: str, y_label: str,
    data: pd.Series, dataset: str, column: str,

  5. Classes (for example the portfolio class) should hold the relevant data and perform no other calculations, these calculations should be done in an independent function.

    • Why? Two reasons:

    These calculations can then be used outside of the class with custom data; for example via the sdk or for tests.

    from openbb_terminal.portfolio.portfolio_helper import get_gaintopain_ratio
    
    # Direct function access
    get_gaintopain_ratio(historical_trade_data, benchmark_trades, benchmark_returns)

    The function can be loaded in SDK factory as an endpoint and user can get result by passing the class instance.

    from openbb_terminal.sdk import openbb
    from openbb_terminal.sdk import Portfolio
    
    transactions = Portfolio.read_orderbook("../../portfolio/holdings/example.csv")
    P = Portfolio(transactions)
    P.generate_portfolio_data()
    P.set_benchmark()
    
    # SDK endpoint access
    openbb.portfolio.gaintopain(P)
    Good code ✅ Bad code ❌
    def get_gaintopain_ratio(portfolio: PortfolioEngine) -> pd.DataFrame:
    
    """..."""
    
    gtp_period_df = portfolio_helper.get_gaintopain_ratio(
      portfolio.historical_trade_data,
      portfolio.benchmark_trades,
      portfolio.benchmark_returns)
    
    return gtp_period_df
    def get_gaintopain_ratio(self) -> pd.DataFrame:
    
    """..."""
    
    vals = list()
    
    for period in portfolio_helper.PERIODS:
           port_rets = portfolio_helper.filter_df_by_period(self.portfolio_returns, period)
           bench_rets =  portfolio_helper.filter_df_by_period(self.benchmark_returns, period)
    
    ...

  6. Naming among related model and view functions should be obvious; just different prefix if possible

    • Why? Eases SDK factory mapping and keeps code clean.

    Good code ✅ Bad code ❌
    # [fred_view.py]
    
    def display_yieldcurve(country: str):
    
          df = fred_model.get_yieldcurve(country)
    
          …
    
    # [fred_model.py]
    
    def get_yieldcurve(country: str) -> pd.Dataframe:
    
          …
    # [fred_view.py]
    
    def display_bondscrv(country: str):
    
          df = fred_model.get_yieldcurve(country)
    
          …
    
    # [fred_model.py]
    
    def get_yldcurve(country: str) -> pd.Dataframe:
    
          …

File Specific Requirements

  1. No data altering in the view file or controller file (view and model with same args)

    • Why? Consistency and good code structure. This also improves the SDK user experience. Thus follows that view and model files will have the same arguments (except for output options like raw, export, external_axes), since no data changes shall be done in the view file.

  2. Each model (get_) should almost always have its own view function (display_)

    • Why? To respect the principles laid out in Code Structure and the previous bullet point. If your code does not have this get_display_ map it’s likely that i. and/or ii. fail to hold.

      i. Data is processed in _model files and displayed in _view files

      ii. _view and _model files will have the same arguments (except for output options)


Coding Style

When in doubt, follow https://www.python.org/dev/peps/pep-0008/.

OpenBB Style Guide

The style guide is a reverse dictionary for argument names, where a brief definition is mapped to an OpenBB recommended argument name and type. When helpful a code snippet example is added below. Following this guide will help keep argument naming consistent and improve SDK users experience.

Style guide structure:

<definition> : <argument_name (argument_type)> e.g. <examples>

def func(..., argument_name: argument_type = default, ...):
    ...

Flags

Show raw data : raw (bool)

def display_data(..., raw: bool = False, ...):
    ...
    if raw:
        print_rich_table(...)

Sort in ascending order : ascend (bool)

def display_data(..., sortby: str = "", ascend: bool = False, ...):
    ...
    if sortby:
        data = data.sort_values(by=sortby, ascend=ascend)

Show plot : plot (bool)

def display_data(..., plot: bool = False, ...):
    ...
    if plot:
        ...
        fig.add_scatter(...)

Output format

Format to export data : export (str), e.g. csv, json, xlsx

def display_data(..., export: str = "", ...):
    ...
    export_data(export, os.path.dirname(os.path.abspath(__file__)), "func", data)

Whether to display plot or return figure (False: display, True: return) : external_axes (bool)

def display_data(..., external_axes: bool = False, ...):
    ...
    fig = OpenBBFigure()
    fig.add_scatter(...)
    return fig.show(external=external_axes)

Field by which to sort : sortby (str), e.g. "Volume"

def display_data(..., sortby: str = "col", ...):
    ...
    if sortby:
        data = data.sort_values(by=sortby)

Maximum limit number of output items : limit (int)

def display_data(..., limit = 10, ...):
    ...
    print_rich_table(
        data[:limit],
        ...
    )

Time-related

Date from which data is fetched (YYYY-MM-DD) : start_date (str), e.g. 2022-01-01

Date up to which data is fetched (YYYY-MM-DD) : end_date (str), e.g. 2022-12-31

Note: We want to accept dates in string because it is easier to deal from user standpoint. Inside the function you can convert it to datetime and check its validity. Please specify date format in docstring.

def get_historical_data(..., start_date: str = "2022-01-01", end_date: str = "2022-12-31",):
    """
    ...
    Parameters
    ----------
    start_date: str
        Date from which data is fetched in format YYYY-MM-DD
    end_date: str
        Date up to which data is fetched in format YYYY-MM-DD
    ...
    """
    data = source_model.get_data(data_name, start_date, end_date, ...)

Year from which data is fetched (YYYY) : start_year (str), e.g. 2022

Year up to which data is fetched (YYYY) : end_year (str), e.g. 2023

def get_historical_data(..., start_year: str = "2022", end_year str = "2023", ...):
    ...
    data = source_model.get_data(data_name, start_year, end_year, ...)

Interval for data observations : interval (str), e.g. 60m, 90m, 1h

def get_prices(interval: str = "60m", ...):
    ...
    data = source.download(
        ...,
        interval=interval,
        ...
    )

Rolling window length : window (int/str), e.g. 252, 252d

def get_rolling_sum(returns: pd.Series, window: str = "252d"):
    rolling_sum = returns.rolling(window=window).sum()

Data selection and manipulation

Search term used to query : query (str)

Maximum limit of search items/periods in data source: limit (int)

Note: please specify limit application in docstring

def get_data_from_source(..., limit: int = 10, ...):
    """
    Parameters
    ----------
    ...
    limit: int
        Number of results to fetch from source
    ...
    """
    data = source.get_data(data_name, n_results=limit, ...)

Dictionary of input datasets : datasets (Dict[str, pd.DataFrame])

Note: Most occurrences are on the econometrics menu and might be refactored in near future

Input dataset : data (pd.DataFrame)

def process_data(..., data: pd.DataFrame, ...):
    """
    ...
    Parameters
    ----------
    ...
    data : pd.DataFrame
        Dataframe of ...
    ...
    """
    col_data = pd.DataFrame(data["Col"])

Dataset name : dataset_name (str)

Input series : data (pd.Series)

Dependent variable series : dependent_series (pd.Series)

Independent variable series : independent_series (pd.Series)

def get_econometric_test(dependent_series, independent_series, ...):
    ...
    dataset = pd.concat([dependent_series, independent_series], axis=1)
    result = econometric_test(dataset, ...)

Country name : country (str), e.g. United States, Portugal

Country initials or abbreviation : country_code (str) e.g. US, PT, USA, POR

Currency to convert data : currency (str) e.g. EUR, USD


Financial instrument characteristics

Instrument ticker, name or currency pair : symbol (str), e.g. AAPL, ethereum, ETH, ETH-USD

def get_prices(symbol: str = "AAPL", ...):
    ...
    data = source.download(
        tickers=symbol,
        ...
    )

Instrument name: name (str)

Note: If a function has both name and symbol as parameter, we should distinguish them and call it name

List of instrument tickers, names or currency pairs : symbols (List/List[str]), e.g. ["AAPL", "MSFT"]

Base currency under BASE-QUOTE → XXX-YYY convention : from_symbol (str), e.g. ETH in ETH-USD

Quote currency under BASE-QUOTE → XXX-YYY convention : to_symbol (str), e.g. USD in ETH-USD

def get_exchange_rate(from_symbol: str = "", to_symbol: str = "", ...):
    ...
    df = source.get_quotes(from_symbol, to_symbol, ...)

Instrument price : price (float)

Instrument implied volatility : implied_volatility (float)

Option strike price : strike_price (float)

Option days until expiration : time_to_expiration (float/str)

Risk free rate : risk_free_rate (float)

Options expiry date : expiry (str)


Naming Convention

  • The name of the variables must be descriptive of what they stand for. I.e. ticker is descriptive, aux is not.
  • Single character variables must be avoided. Except if they correspond to the iterator of a loop.

Docstrings

The docstring format used in numpy, an example is shown below:

def command_foo(var1: str, var2: List[int], var3: bool = False) -> Tuple[int, pd.DataFrame]:
"""Small description

[Optional: Longer description]

Parameters
----------
var1 : str
    var1 description
var2 : List[int]
    var2 description
var3 : bool, optional
    var3 description

Returns
-------
foo : int
    returned foo description
pd.DataFrame
    dataframe returned
"""

Linters

The following linters are used by our codebase:

Linter Description
bandit security analyzer
black code formatter
codespell spelling checker
ruff a fast python linter
mypy static typing checker
pylint static code analysis
markdownlint markdown linter

Command names

  • The command name should be as short as possible.

  • The command name should allow the user to know what the command refers to without needing to read description. (e.g. earn)

    • If this is not possible, then the command name should be an abbreviation of what the functionality corresponds to (e.g. ycrv for yield curve)
  • The command name should not have the data source explicit

UI and UX

Screenshot 2022-10-26 at 12 17 19

It is important to keep a coherent UI/UX throughout the terminal. These are the rules we must abide:

  • There is 1 single empty line between user input and start of the command output.
  • There is 1 single empty line between command output and the user input.
  • The menu help has 1 empty line above text and 1 empty line below. Both still within the rectangular panel.
  • From menu help rectangular panel there's no empty line below - this makes it more clear to the user that they are inside such menu.

External API Keys

Creating API key

OpenBB Terminal currently has over 100 different data sources. Most of these require an API key that allows access to some free tier features from the data provider, but also paid ones.

When a new API data source is added to the platform, it must be added through credentials_model.py.

In order to do that, you'll simply need to choose from one the following files:

  1. local_credentials.json --> credentials that should only be stored locally and not pushed to the OpenBB Hub like brokerage keys or other very sensitive or personal to the user.
  2. hub_credentials.json --> credentials that should be stored in the OpenBB Hub like API keys to access your favorite providers.

Then just update all_api_keys.json with the instructions to get the api key from the data source website. Make sure that this file has the correct .json format, otherwise the API keys page in the Hub will break (e.g. in json the last element key-value pair shouldn't be followed by a comma, and the last object in a list of dictionaries should also not be followed by a comma).

Note: By differentiating between local and hub credentials, we can ensure that the user's credentials are not pushed to the OpenBB Hub and are only stored locally. This does not mean that the credentials are not secure in the OpenBB Hub, but rather that the user can choose to store them locally if they wish.

Setting and checking API key

One of the first steps once adding a new data source that requires an API key is to add that key to our keys_controller.py. This menu allows the user to set API keys and check their validity.

The following code allows to check the validity of the Polygon API key.

def check_polygon_key(show_output: bool = False) -> str:
    """Check Polygon key

    Parameters
    ----------
    show_output: bool
        Display status string or not. By default, False.

    Returns
    -------
    str
        Status of key set
    """

    current_user = get_current_user()

    if current_user.credentials.API_POLYGON_KEY == "REPLACE_ME":
        logger.info("Polygon key not defined")
        status = KeyStatus.NOT_DEFINED
    else:
        r = request(
            "https://api.polygon.io/v2/aggs/ticker/AAPL/range/1/day/2020-06-01/2020-06-17"
            f"?apiKey={current_user.credentials.API_POLYGON_KEY}"
        )
        if r.status_code in [403, 401]:
            logger.warning("Polygon key defined, test failed")
            status = KeyStatus.DEFINED_TEST_FAILED
        elif r.status_code == 200:
            logger.info("Polygon key defined, test passed")
            status = KeyStatus.DEFINED_TEST_PASSED
        else:
            logger.warning("Polygon key defined, test inconclusive")
            status = KeyStatus.DEFINED_TEST_INCONCLUSIVE

    if show_output:
        console.print(status.colorize())

    return str(status)

Note that there are usually 3 states:

  • defined, test passed: The user has set their API key and it is valid.
  • defined, test failed: The user has set their API key but it is not valid.
  • not defined: The user has not defined any API key.

Note: Sometimes the user may have the correct API key but still not have access to a feature from that data source, and that may be because such feature required an API key of a higher level.

A function can then be created with the following format to allow the user to change its environment key directly from the terminal.

def call_polygon(self, other_args: List[str]):
    """Process polygon command"""
    parser = argparse.ArgumentParser(
        add_help=False,
        formatter_class=argparse.ArgumentDefaultsHelpFormatter,
        prog="polygon",
        description="Set Polygon API key.",
    )
    parser.add_argument(
        "-k",
        "--key",
        type=str,
        dest="key",
        help="key",
    )
    if not other_args:
        console.print("For your API Key, visit: https://polygon.io")
        return

    if other_args and "-" not in other_args[0][0]:
        other_args.insert(0, "-k")
    ns_parser = self.parse_simple_args(parser, other_args)
    if ns_parser:
        self.status_dict["polygon"] = keys_model.set_polygon_key(
            key=ns_parser.key, persist=True, show_output=True
        )

ADVANCED

Important functions and classes

Base controller class

This BaseController class is inherited by all controllers on the terminal.

This class contains both important variables and methods that are common across all terminal controllers.

CHOICES_COMMON: List of common commands across all controllers

  • cls: clear screen
  • home: go back to the main root
  • about: allows to open our documentation directly on the menu or command
  • h, ? and help: display the help menu the user is in
  • q, quit and ..: go back to one menu above
  • exit: exit the platform
  • r and reset: reset the platform (reading code and settings again but going into the same state)
  • support: create a support request ticket

All of these variables have a call_FUNCTION associated with them.

Worthy methods to mention are:

  • load_class: Checks for an existing instance of the controller before creating a new one to speed up access to that menu.
  • custom_reset: Should be used by controllers that rely on a state variable - meant to be overridden. They should add the commands necessary to have the same data loaded.
  • print_help: Meant to be overridden by each controller
  • parse_input: Processes the string the user inputs into a list of actionable commands
  • switch: Acts upon the command action received
  • parse_known_args_and_warn: Parses the command with the - and -- flags and variables. Some built-in flags are:
    • export_allowed: Which can be set to _NO_EXPORT_, _EXPORT_ONLY_RAW_DATA_ALLOWED_, _EXPORT_ONLY_FIGURES_ALLOWED_ and _EXPORT_BOTH_RAW_DATA_AND_FIGURES_
    • raw: Displaying the data raw
    • limit: Number of rows to display
  • menu: Most important method. When a menu is executed, the way to call it is through stocks_menu.menu()

Default Data Sources

The document openbb_default.json contains all data sources that the terminal has access to and specifies the data source utilized by default for each command.

The convention is as follows:

{
  "stocks": {
    "search": [
      "FinanceDatabase"
    ],
    "quote": [
      "FinancialModelingPrep"
    ],
    "tob": [
      "CBOE"
    ],
    "candle": [],
    "codes": [
      "Polygon"
    ],
    "news": [
      "Feedparser",
      "NewsApi",
    ],
            ...

The way to interpret this file is by following the path to a data source, e.g.

  • stocks/search relies on FinanceDatabase
  • stocks/candle does not rely on any data source. This means that it relies on data that has been loaded before.
  • stocks/load relies on YahooFinance, AlphaVantage, Polygon or EODHD.
    • The order is important as the first data source is the one utilized by default.
  • stocks/options/unu relies on FDScanner.
  • stocks/options/exp relies on YahooFinance by default but Tradier and Nasdaq sources are allowed.

Note: The default data sources can be changed directly in the OpenBB Hub by the user and automatically synchronized with the terminal on login.

Export Data

In the _view.py files it is common having at the end of each function export_data being called. This typically looks like:

    export_data(
        export,
        os.path.dirname(os.path.abspath(__file__)),
        "pt",
        df_analyst_data,
        sheet_name,
        fig,
    )

Let's go into each of these arguments:

  • export corresponds to the type of file we are exporting.
    • If the user doesn't have anything selected, then this function doesn't do anything.
    • The user can export multiple files and even name the files.
    • The allowed type of files json,csv,xlsx for raw data and jpg,pdf,png,svg for figures depends on the export_allowed variable defined in parse_known_args_and_warn.
  • os.path.dirname(os.path.abspath(__file__)) corresponds to the directory path
    • This is important when export folder selected is the default because the data gets stored based on where it is called.
    • If this is called from a common folder, we can use os.path.dirname(os.path.abspath(__file__)).replace("common", "stocks") instead
  • "pt" corresponds to the name of the exported file (+ unique datetime) if the user doesn't provide one
  • df_analyst_data corresponds to the dataframe with data.
  • sheet_name corresponds to the name of the sheet in the excel file.
  • fig corresponds to the figure to be exported as an image or pdf.

If export_allowed=EXPORT_BOTH_RAW_DATA_AND_FIGURES in parse_known_args_and_warn, valid examples are:

  • cmd --export csv
  • cmd --export csv,png,jpg
  • cmd --export mydata.csv
  • cmd --export mydata.txt,alsomydata.csv,alsoalsomydata.png

Note that these files are saved on a location based on the environment variable: EXPORT_FOLDER_PATH. Which can be set in settings/export.

The default location is the exports folder and the data will be stored with the same organization of the terminal. But, if the user specifies the name of the file, then that will be dropped onto the folder as is with the datetime attached.

Queue and pipeline

The variable self.queue contains a list of all actions to be run on the platform. That is the reason why this variable is always passed as an argument to a new controller class and received back.

  self.queue = self.load_class(
      DarkPoolShortsController, self.ticker, self.start, self.stock, self.queue
  )

Example:

If a user is in the root of the terminal and runs:

stocks/load AAPL/dps/psi -l 90

The queue created becomes: self.queue = ["stocks", "load AAPL", "dps", "psi -l 90"]

And the user goes into the stocks menu and runs load AAPL. Then the queue is updated to self.queue = ["dps", "psi -l 90"]

At that point the user goes into the dps menu and runs the command psi with the argument -l 90 therefore displaying price vs short interest of the past 90 days.

Auto Completer

In order to help users with a powerful autocomplete, we have implemented our own (which can be found here).

The queue, discussed in the previous section Queue and pipeline, is expected to link together with the autocompletion in order to provide the user with the available options for each command. Here is an example of how it will look like:

2023 Apr 11, 11:41 (🦋) /stocks/dps/ $ psi
                                            --nyse
                                            --help
                                            -h
                                            --export
                                            --raw
                                            --limit
                                            -l
                                            --source

Where nyse, help, h, export, raw, limit, l and source are the available options for the psi command. Those are selectable using the arrow keys and the tab key.

The list of options for each command is automatically generated, if you're interested take a look at its implementation here.

To leverage this functionality, you need to add the following line to the top of the desired controller:

CHOICES_GENERATION = True

Here's an example of how to use it, on the forex controller:

class ForexController(BaseController):
    """Forex Controller class."""

    CHOICES_COMMANDS = [
        "fwd",
        "candle",
        "load",
        "quote",
    ]
    CHOICES_MENUS = [
        "forecast",
        "qa",
        "ta",
    ]
    RESOLUTION = ["i", "d", "w", "m"]

    PATH = "/forex/"
    FILE_PATH = os.path.join(os.path.dirname(__file__), "README.md")
    CHOICES_GENERATION = True

    def __init__(self, queue: Optional[List[str]] = None):
        """Construct Data."""
        super().__init__(queue)

        self.fx_pair = ""
        self.from_symbol = ""
        self.to_symbol = ""
        self.source = get_ordered_list_sources(f"{self.PATH}load")[0]
        self.data = pd.DataFrame()

        if session and get_current_user().preferences.USE_PROMPT_TOOLKIT:
            choices: dict = self.choices_default
            choices["load"].update({c: {} for c in FX_TICKERS})

            self.completer = NestedCompleter.from_nested_dict(choices)


    ...

In case the user is interested in a DYNAMIC list of options which changes based on user's state, then a class method must be defined.

The example below shows an excerpt from update_runtime_choices method in the options controller.

def update_runtime_choices(self):
        """Update runtime choices"""
        if session and get_current_user().preferences.USE_PROMPT_TOOLKIT:
            if not self.chain.empty:
                strike = set(self.chain["strike"])

                self.choices["hist"]["--strike"] = {str(c): {} for c in strike}
                self.choices["grhist"]["-s"] = "--strike"
                self.choices["grhist"]["--strike"] = {str(c): {} for c in strike}
                self.choices["grhist"]["-s"] = "--strike"
                self.choices["binom"]["--strike"] = {str(c): {} for c in strike}
                self.choices["binom"]["-s"] = "--strike"

This method should only be called when the user's state changes leads to the auto-complete not being accurate.

In this case, this method is called as soon as the user successfully loads a new ticker since the options expiry dates vary based on the ticker. Note that the completer is recreated from it.

Logging

A logging system is used to help tracking errors inside the OpenBBTerminal.

This is storing every logged message inside the following location :

$HOME/OpenBBUserData/logs

Where $HOME is the user home directory, for instance:

  • C:\Users\foo if you are in Windows and your name is foo
  • /home/bar/ if you are in macOS or Linux and your name is bar

The user can override this location using the settings key OPENBB_USER_DATA_DIRECTORY.

If you want to log a particular message inside a function you can do like so:

import logging

logger = logging.getLogger(__name__)

def your_function() -> pd.DataFrame:
    logger.info("Some log message with the level INFO")
    logger.warning("Some log message with the level WARNING")
    logger.fatal("Some log message with the level FATAL")

You can also use the decorator @log_start_end to automatically record a message every time a function starts and ends, like this:

import logging

from openbb_terminal.decorators import log_start_end

logger = logging.getLogger(__name__)

@log_start_end(log=logger)
def your_function() -> pd.DataFrame:
    pass

Note: if you don't want your logs to be collected, you can set the OPENBB_LOG_COLLECT environment variable on your .env file to False.

Disclaimer: all the user paths, names, IPs, credentials and other sensitive information are anonymized, take a look at how we do it.

Internationalization

WORK IN PROGRESS - The menu can be internationalised BUT we do not support yet help commands-h internationalization.

In order to add support for a new language, the best approach is to:

  1. Copy-paste i18n/en.yml
  2. Rename that file to a short version of language you are translating to, e.g. i18n/pt.yml for portuguese
  3. Then just update the text on the right. E.g.
  stocks/NEWS: latest news of the company

becomes

  stocks/NEWS: mais recentes notícias da empresa

Note: To speed up translation, the team developed a script that uses Google translator API to help translating the entire en.yml document to the language of choice. Then the output still needs to be reviewed, but this can be a useful bootstrap.

This is the convention in use for creating a new key/value pair:

  • stocks/search - Under stocks context, short command search description on the help menu
  • stocks/SEARCH - Under stocks context, long command search description, when search -h
  • stocks/SEARCH_query - Under stocks context, query description when inquiring about search command with search -h
  • stocks/_ticker - Under stocks context, _ticker is used as a key of a parameter, and the displayed parameter description is given as value
  • crypto/dd/_tokenomics_ - Under crypto context and under dd menu, _tokenomics_ is used as a key of an additional information, and the displayed information is given as value

Settings

The majority of the settings used in the OpenBB Terminal are handled using Pydantic Dataclasses. Some examples are:

  1. SystemModel
  2. UserModel
  3. CredentialsModel
  4. ...

This means that the settings are pretty much validated and documented automatically, as well as centralized in a single place. This allows us to develop faster, efficiantly and with predictability. Depending on your use case you'll most likely need to interact with these dataclasses or expand them with new settings.

Disclaimer: avoid at all costs to use the os.environ or os.getenv methods to retrieve settings. Settings should be retrieved using the appropriate methods from the respective class.

Here is an example of accessing a setting:

from openbb_terminal.core.session.current_system import get_current_system

system = get_current_system()
system = get_current_system()
print(system.VERSION)

# 3.0.0

And here is an example of changing a setting:

from openbb_terminal.core.session.current_system import get_current_system

set_system_variable("TEST_MODE", True)

Write Code and Commit

At this stage it is assumed that you have already forked the project and are ready to start working.

Pre Commit Hooks

Git hook scripts are useful for identifying simple issues before submission to code review. We run our hooks on every commit to automatically point out issues in code such as missing semicolons, trailing whitespace, and debug statements. By pointing these issues out before code review, this allows a code reviewer to focus on the architecture of a change while not wasting time with trivial style nitpicks.

Install the pre-commit hooks by running: pre-commit install.

Coding

Although the Coding Guidelines section has been already explained, it is worth mentioning that if you want to be faster at developing a new feature, you may implement it first on a jupyter notebook and then carry it across to the terminal. This is mostly useful when the feature relies on scraping data from a website, or implementing a Neural Network model.

Git Process

  1. Create your Feature Branch, e.g. git checkout -b feature/AmazingFeature
  2. Check the files you have touched using git status
  3. Stage the files you want to commit, e.g. git add openbb_terminal/stocks/stocks_controller.py openbb_terminal/stocks/stocks_helper.py. Note: DON'T add config_terminal.py or .env files with personal information, or even feature_flags.py which is user-dependent.
  4. Write a concise commit message under 50 characters, e.g. git commit -m "meaningful commit message". If your PR solves an issue raised by a user, you may specify such issue by adding #ISSUE_NUMBER to the commit message, so that these get linked. Note: If you installed pre-commit hooks and one of the formatters re-formats your code, you'll need to go back to step 3 to add these.

Branch Naming Conventions

The accepted branch naming conventions are:

  • feature/feature-name
  • hotfix/hotfix-name
  • release/2.1.0 or release/2.1.0rc0.
  • bugfix/bugfix-name
  • docs/docs-name

All feature/feature-name and bugfix/bugfix-name related branches can only have PRs pointing to develop branch. release/* branches can only have PRs pointing to main branch, while hotfix/hotfix-name should first be merged to main and then into develop to sync the hotfix changes.

When develop branch is merged to main, a GitHub action will run scripts that generate documentation content (reference sections, data models, etc.) and trigger the website deployment. Those scripts can be found in the following path website/generate_*.py.

The develop branch is only merged to main right before a new release, but sometimes you might need to update the website in-between releases. To do this follow these steps:

  1. create docs/[my-update] branch from main
  2. commit your changes to docs/[my-update]
  3. merge docs/[my-update] into main -> website deployment triggered
  4. merge docs/[my-update] into develop -> NO website deployment, just to sync branches

Installers

When implementing a new feature or fixing something within the codebase, it is necessary to ensure that it is working appropriately on the terminal. However, it is equally as important to ensure that new features or fixes work on the installer terminal too. This is because a large portion of users utilize the installer to use OpenBB Terminal. More information on how to build an installer can be found here.