sqlglot.lineage

  1from __future__ import annotations
  2
  3import json
  4import typing as t
  5from dataclasses import dataclass, field
  6
  7from sqlglot import Schema, exp, maybe_parse
  8from sqlglot.errors import SqlglotError
  9from sqlglot.optimizer import Scope, build_scope, qualify
 10
 11if t.TYPE_CHECKING:
 12    from sqlglot.dialects.dialect import DialectType
 13
 14
 15@dataclass(frozen=True)
 16class Node:
 17    name: str
 18    expression: exp.Expression
 19    source: exp.Expression
 20    downstream: t.List[Node] = field(default_factory=list)
 21    alias: str = ""
 22
 23    def walk(self) -> t.Iterator[Node]:
 24        yield self
 25
 26        for d in self.downstream:
 27            if isinstance(d, Node):
 28                yield from d.walk()
 29            else:
 30                yield d
 31
 32    def to_html(self, **opts) -> LineageHTML:
 33        return LineageHTML(self, **opts)
 34
 35
 36def lineage(
 37    column: str | exp.Column,
 38    sql: str | exp.Expression,
 39    schema: t.Optional[t.Dict | Schema] = None,
 40    sources: t.Optional[t.Dict[str, str | exp.Subqueryable]] = None,
 41    dialect: DialectType = None,
 42    **kwargs,
 43) -> Node:
 44    """Build the lineage graph for a column of a SQL query.
 45
 46    Args:
 47        column: The column to build the lineage for.
 48        sql: The SQL string or expression.
 49        schema: The schema of tables.
 50        sources: A mapping of queries which will be used to continue building lineage.
 51        dialect: The dialect of input SQL.
 52        **kwargs: Qualification optimizer kwargs.
 53
 54    Returns:
 55        A lineage node.
 56    """
 57
 58    expression = maybe_parse(sql, dialect=dialect)
 59
 60    if sources:
 61        expression = exp.expand(
 62            expression,
 63            {
 64                k: t.cast(exp.Subqueryable, maybe_parse(v, dialect=dialect))
 65                for k, v in sources.items()
 66            },
 67        )
 68
 69    qualified = qualify.qualify(
 70        expression,
 71        dialect=dialect,
 72        schema=schema,
 73        **{"validate_qualify_columns": False, "identify": False, **kwargs},  # type: ignore
 74    )
 75
 76    scope = build_scope(qualified)
 77
 78    if not scope:
 79        raise SqlglotError("Cannot build lineage, sql must be SELECT")
 80
 81    def to_node(
 82        column: str | int,
 83        scope: Scope,
 84        scope_name: t.Optional[str] = None,
 85        upstream: t.Optional[Node] = None,
 86        alias: t.Optional[str] = None,
 87    ) -> Node:
 88        aliases = {
 89            dt.alias: dt.comments[0].split()[1]
 90            for dt in scope.derived_tables
 91            if dt.comments and dt.comments[0].startswith("source: ")
 92        }
 93
 94        # Find the specific select clause that is the source of the column we want.
 95        # This can either be a specific, named select or a generic `*` clause.
 96        select = (
 97            scope.expression.selects[column]
 98            if isinstance(column, int)
 99            else next(
100                (select for select in scope.expression.selects if select.alias_or_name == column),
101                exp.Star() if scope.expression.is_star else None,
102            )
103        )
104
105        if not select:
106            raise ValueError(f"Could not find {column} in {scope.expression}")
107
108        if isinstance(scope.expression, exp.Union):
109            upstream = upstream or Node(name="UNION", source=scope.expression, expression=select)
110
111            index = (
112                column
113                if isinstance(column, int)
114                else next(
115                    i
116                    for i, select in enumerate(scope.expression.selects)
117                    if select.alias_or_name == column
118                )
119            )
120
121            for s in scope.union_scopes:
122                to_node(index, scope=s, upstream=upstream)
123
124            return upstream
125
126        if isinstance(scope.expression, exp.Select):
127            # For better ergonomics in our node labels, replace the full select with
128            # a version that has only the column we care about.
129            #   "x", SELECT x, y FROM foo
130            #     => "x", SELECT x FROM foo
131            source = t.cast(exp.Expression, scope.expression.select(select, append=False))
132        else:
133            source = scope.expression
134
135        # Create the node for this step in the lineage chain, and attach it to the previous one.
136        node = Node(
137            name=f"{scope_name}.{column}" if scope_name else str(column),
138            source=source,
139            expression=select,
140            alias=alias or "",
141        )
142        if upstream:
143            upstream.downstream.append(node)
144
145        # Find all columns that went into creating this one to list their lineage nodes.
146        for c in set(select.find_all(exp.Column)):
147            table = c.table
148            source = scope.sources.get(table)
149
150            if isinstance(source, Scope):
151                # The table itself came from a more specific scope. Recurse into that one using the unaliased column name.
152                to_node(
153                    c.name, scope=source, scope_name=table, upstream=node, alias=aliases.get(table)
154                )
155            else:
156                # The source is not a scope - we've reached the end of the line. At this point, if a source is not found
157                # it means this column's lineage is unknown. This can happen if the definition of a source used in a query
158                # is not passed into the `sources` map.
159                source = source or exp.Placeholder()
160                node.downstream.append(Node(name=c.sql(), source=source, expression=source))
161
162        return node
163
164    return to_node(column if isinstance(column, str) else column.name, scope)
165
166
167class LineageHTML:
168    """Node to HTML generator using vis.js.
169
170    https://visjs.github.io/vis-network/docs/network/
171    """
172
173    def __init__(
174        self,
175        node: Node,
176        dialect: DialectType = None,
177        imports: bool = True,
178        **opts: t.Any,
179    ):
180        self.node = node
181        self.imports = imports
182
183        self.options = {
184            "height": "500px",
185            "width": "100%",
186            "layout": {
187                "hierarchical": {
188                    "enabled": True,
189                    "nodeSpacing": 200,
190                    "sortMethod": "directed",
191                },
192            },
193            "interaction": {
194                "dragNodes": False,
195                "selectable": False,
196            },
197            "physics": {
198                "enabled": False,
199            },
200            "edges": {
201                "arrows": "to",
202            },
203            "nodes": {
204                "font": "20px monaco",
205                "shape": "box",
206                "widthConstraint": {
207                    "maximum": 300,
208                },
209            },
210            **opts,
211        }
212
213        self.nodes = {}
214        self.edges = []
215
216        for node in node.walk():
217            if isinstance(node.expression, exp.Table):
218                label = f"FROM {node.expression.this}"
219                title = f"<pre>SELECT {node.name} FROM {node.expression.this}</pre>"
220                group = 1
221            else:
222                label = node.expression.sql(pretty=True, dialect=dialect)
223                source = node.source.transform(
224                    lambda n: exp.Tag(this=n, prefix="<b>", postfix="</b>")
225                    if n is node.expression
226                    else n,
227                    copy=False,
228                ).sql(pretty=True, dialect=dialect)
229                title = f"<pre>{source}</pre>"
230                group = 0
231
232            node_id = id(node)
233
234            self.nodes[node_id] = {
235                "id": node_id,
236                "label": label,
237                "title": title,
238                "group": group,
239            }
240
241            for d in node.downstream:
242                self.edges.append({"from": node_id, "to": id(d)})
243
244    def __str__(self):
245        nodes = json.dumps(list(self.nodes.values()))
246        edges = json.dumps(self.edges)
247        options = json.dumps(self.options)
248        imports = (
249            """<script type="text/javascript" src="https://unpkg.com/vis-data@latest/peer/umd/vis-data.min.js"></script>
250  <script type="text/javascript" src="https://unpkg.com/vis-network@latest/peer/umd/vis-network.min.js"></script>
251  <link rel="stylesheet" type="text/css" href="https://unpkg.com/vis-network/styles/vis-network.min.css" />"""
252            if self.imports
253            else ""
254        )
255
256        return f"""<div>
257  <div id="sqlglot-lineage"></div>
258  {imports}
259  <script type="text/javascript">
260    var nodes = new vis.DataSet({nodes})
261    nodes.forEach(row => row["title"] = new DOMParser().parseFromString(row["title"], "text/html").body.childNodes[0])
262
263    new vis.Network(
264        document.getElementById("sqlglot-lineage"),
265        {{
266            nodes: nodes,
267            edges: new vis.DataSet({edges})
268        }},
269        {options},
270    )
271  </script>
272</div>"""
273
274    def _repr_html_(self) -> str:
275        return self.__str__()