Export

The reclink.export module provides lightweight helpers for persisting match results and cluster assignments to disk. All functions use the Python standard library only -- no extra dependencies required.

from reclink.export import (
    export_matches_csv,
    export_matches_json,
    export_clusters_csv,
    export_clusters_json,
)

:::tip Input flexibility Every export function accepts a plain list[dict], a Pandas DataFrame, or a Polars DataFrame. The format is detected automatically so you can pass pipeline output directly. :::

---

Match Export

Match results are rows containing left_id, right_id, score, and an optional scores list of per-field scores.

export_matches_csv(results, path)

Write match results to a CSV file. The scores list is serialized as a semicolon-delimited string.

from reclink.export import export_matches_csv

matches = [
    {"left_id": "0", "right_id": "3", "score": 0.92, "scores": [0.95, 0.88]},
    {"left_id": "1", "right_id": "4", "score": 0.87, "scores": [0.90, 0.84]},
]

export_matches_csv(matches, "matches.csv")

The resulting CSV:

left_id,right_id,score,scores
0,3,0.92,0.95;0.88
1,4,0.87,0.9;0.84

export_matches_json(results, path)

Write match results to a JSON file with 2-space indentation.

from reclink.export import export_matches_json

export_matches_json(matches, "matches.json")

The resulting JSON:

[
  {
    "left_id": "0",
    "right_id": "3",
    "score": 0.92,
    "scores": [0.95, 0.88]
  }
]

Parameters

Parameter	Type	Default	Description
results	list[dict], pd.DataFrame, or pl.DataFrame	required	Match results. Each row must contain left_id, right_id, and score keys. An optional scores list holds per-field scores.
path	str or Path	required	Output file path. Parent directories must exist.

---

Cluster Export

Cluster data can be either a list of groups (each group is a list of record IDs) or a DataFrame with cluster_id and record_id columns.

export_clusters_csv(clusters, path)

Write cluster assignments to CSV with cluster_id and record_id columns. When given a list of groups, cluster IDs are assigned sequentially starting from 0.

from reclink.export import export_clusters_csv

clusters = [
    ["rec_0", "rec_3", "rec_7"],   # cluster 0
    ["rec_1", "rec_4"],            # cluster 1
]

export_clusters_csv(clusters, "clusters.csv")

The resulting CSV:

cluster_id,record_id
0,rec_0
0,rec_3
0,rec_7
1,rec_1
1,rec_4

export_clusters_json(clusters, path)

Write cluster assignments to JSON. Each entry is an object with cluster_id and record_id fields.

from reclink.export import export_clusters_json

export_clusters_json(clusters, "clusters.json")

The resulting JSON:

[
  {"cluster_id": 0, "record_id": "rec_0"},
  {"cluster_id": 0, "record_id": "rec_3"},
  {"cluster_id": 0, "record_id": "rec_7"},
  {"cluster_id": 1, "record_id": "rec_1"},
  {"cluster_id": 1, "record_id": "rec_4"}
]

Parameters

Parameter	Type	Default	Description
clusters	list[list[str]], pd.DataFrame, or pl.DataFrame	required	Cluster groups. Lists are auto-indexed; DataFrames must contain cluster_id and record_id columns.
path	str or Path	required	Output file path. Parent directories must exist.

---

End-to-End Example

A typical workflow: run a deduplication pipeline, then export both matches and clusters.

import reclink
from reclink.pipeline import ReclinkPipeline
from reclink.export import export_matches_csv, export_clusters_json

# Build and run a dedup pipeline
pipeline = (
    ReclinkPipeline.builder()
    .preprocess("name", ["fold_case", "normalize_whitespace"])
    .compare_string("name", metric="jaro_winkler")
    .classify_threshold(0.85)
    .build()
)
matches = pipeline.dedup(data, id_column="id")

# Export match pairs
export_matches_csv(matches, "output/matches.csv")

# Cluster and export
clusters = pipeline.dedup_cluster(data, id_column="id")
export_clusters_json(clusters, "output/clusters.json")

---

See Also

• Pipeline -- build record linkage pipelines that produce match results
• Evaluation -- measure precision, recall, and F1 of your matches
• CLI Reference -- export results directly from the command line