feat(connector): don't support xml website anymore

All the supported websites return JSON. Moreover, supporting xml
response requires us to add
lxml as the dependency.
We can add this back if we find a website returns xml.

dataprep/connector/implicit_database.py

@@ -3,19 +3,15 @@
 where ImplicitDatabase is a conceptual model describes
 a website and ImplicitTable describes an API endpoint.
 """
-from io import StringIO
 from json import load as jload
 from json import loads as jloads
 from pathlib import Path
 from typing import Any, Dict, List, Union
 
 import pandas as pd
 from jsonpath_ng import parse as jparse
-from lxml import etree  # pytype: disable=import-error
-
 from dataprep.connector.schema.defs import ConfigDef
 
-from ..errors import UnreachableError
 from .schema import ConfigDef
 
 _TYPE_MAPPING = {
@@ -45,10 +41,8 @@ def from_response(self, payload: str) -> pd.DataFrame:
         ctype = self.config.response.ctype  # pylint: disable=no-member
         if ctype == "application/json":
             rows = self.from_json(payload)
-        elif ctype == "application/xml":
-            rows = self.from_xml(payload)
         else:
-            raise UnreachableError
+            raise NotImplementedError(f"{ctype} not supported")
 
         return pd.DataFrame(rows)
 
@@ -98,51 +92,6 @@ def from_json(self, data: str) -> Dict[str, List[Any]]:
 
         return table_data
 
-    def from_xml(self, data: str) -> Dict[str, List[Any]]:
-        """
-        Create rows from xml string.
-        """
-        table_data = {}
-        respdef = self.config.response
-        data = data.replace('<?xml version="1.0" encoding="UTF-8"?>', "")
-
-        root = etree.parse(StringIO(data))
-        data_rows = root.xpath(respdef.table_path)
-
-        if respdef.orient == "records":
-            for (
-                column_name,
-                column_def,
-            ) in respdef.schema_.items():
-                column_target = column_def.target
-                column_type = column_def.type
-
-                col: List[Any] = []
-                for data_row in data_rows:
-                    maybe_cell_value = data_row.xpath(column_target)
-
-                    if not maybe_cell_value:
-                        col.append(None)
-                    elif len(maybe_cell_value) == 1 and column_type != "object":
-                        (cell_value,) = maybe_cell_value
-                        if cell_value is not None:
-                            # Even we have value matched,
-                            # the value might be None so we don't do type conversion.
-                            cell_value = _TYPE_MAPPING[column_type](cell_value)
-                        col.append(cell_value)
-                    else:
-                        assert (
-                            column_type == "object"
-                        ), f"{column_name}: {maybe_cell_value} is not {column_type}"
-                        col.append(maybe_cell_value)
-
-                table_data[column_name] = col
-        else:
-            # TODO: split orient
-            raise NotImplementedError
-
-        return table_data
-
 
 class ImplicitDatabase:
     """

dataprep/connector/schema/defs.py

@@ -370,7 +370,7 @@ def merge_type(a: str, b: str) -> str:  # pylint: disable=invalid-name
 
 
 class ResponseDef(BaseDef):
-    ctype: str = Field(regex=r"^(application/xml|application/json)$")
+    ctype: str = Field("application/json", const=True)
     table_path: str
     schema_: Dict[str, SchemaFieldDef] = Field(alias="schema")
     orient: str = Field(regex=r"^(records|split)$")

pyproject.toml

@@ -51,7 +51,6 @@ bottleneck = "^1.3.2"
 jsonschema = "~3.2"
 requests = "~2.24"
 jinja2 = "~2.11"
-lxml = "~4.5"
 tqdm = "~4.48"
 tornado = "5.0.2"
 jsonpath-ng = "~1.5"