Module tf.browser.ner.web

Flask sub web app in the TF browser.

Expand source code Browse git
"""Flask sub web app in the TF browser.
"""

from flask import Blueprint, send_file

from ...core.generic import AttrDict
from ...core.files import abspath, fileExists, dirNm
from ...ner.settings import TOOLKEY
from ...ner.ner import NER

from .serve import serveNer, serveNerContext


MY_DIR = dirNm(abspath(__file__))

METHODS = ["GET", "POST"]


def factory(web):
    """A sub web app, to be inserted into the TF browser web app.

    The way of connecting this sub app to the main app is by way of the concept
    of [BluePrint](https://flask.palletsprojects.com/en/2.3.x/blueprints/),
    which is built into Flask itself.

    Before starting the actual serving of pages, we initialize a
    `tf.ner.ner.NER` object, and store it under attribute `ner`.

    In order to do so, we pick up a handle to the loaded TF corpus,
    and a handle to the tool data, both present in the `web` object (see parameters
    below).


    Parameters
    ----------
    web: object
        This represents the Flask website that is the TF browser.

        We may assume that an API for a loaded TF corpus is present under attribute
        `kernelApi`.

        Possibly there is also an attribute `toolData`, which is the store for all
        tool specific data.
        If not, we create an empty store. Inside that store we create an empty
        sub-store for this specific tool.
        The initialization of the `NER` object makes sure this store is
        populated by the tool data as it is read from disk.

        This way, the annotation data is preserved between requests.
    """
    app = Blueprint(
        TOOLKEY,
        __name__,
        url_prefix=f"/{TOOLKEY}",
        template_folder="templates",
    )
    kernelApi = web.kernelApi
    tfApp = kernelApi.app

    if not hasattr(web, "toolData"):
        setattr(web, "toolData", AttrDict())

    toolData = web.toolData

    if TOOLKEY not in toolData:
        toolData[TOOLKEY] = AttrDict()

    data = toolData[TOOLKEY]
    web.ner = NER(tfApp, data=data, browse=True)

    if not web.ner.properlySetup:
        return app

    @app.route("/static/<path:filepath>")
    def serveStatic(filepath):
        theFile = f"{MY_DIR}/static/{filepath}"
        return send_file(theFile) if fileExists(theFile) else ""

    @app.route("/index", methods=METHODS)
    def serveNerX():
        return serveNer(web)

    @app.route("/context/<int:node>", methods=METHODS)
    def serveNerContextX(node):
        return serveNerContext(web, node)

    @app.route("/<path:anything>", methods=METHODS)
    def serveAllX(anything=None):
        return f"path={anything}"

    return app

Functions

def factory(web)

A sub web app, to be inserted into the TF browser web app.

The way of connecting this sub app to the main app is by way of the concept of BluePrint, which is built into Flask itself.

Before starting the actual serving of pages, we initialize a NER object, and store it under attribute ner.

In order to do so, we pick up a handle to the loaded TF corpus, and a handle to the tool data, both present in the web object (see parameters below).

Parameters

web : object

This represents the Flask website that is the TF browser.

We may assume that an API for a loaded TF corpus is present under attribute kernelApi.

Possibly there is also an attribute toolData, which is the store for all tool specific data. If not, we create an empty store. Inside that store we create an empty sub-store for this specific tool. The initialization of the NER object makes sure this store is populated by the tool data as it is read from disk.

This way, the annotation data is preserved between requests.