Consistency in usage of sql_dialect, sql_dialect_str, sqlglot_dialect

[RobinL]

This PR addresses inconsistent usage of variable names for SQL dialects across the Splink codebase.

Currently, there is ambiguity between variable names such as sql_dialect, sql_dialect_str, dialect, sqlglot_name, and sqlglot_dialect. The confusion stems from the differences between the SQLGlot library, which uses string representations of SQL dialects (e.g., "spark", "duckdb"), and Splink’s own SplinkDialect, which can be instantiated from strings (e.g., SplinkDialect.from_string("duckdb")). These representations can be inconsistent, such as "athena" in Splink being "presto" in SQLGlot.

This PR ensures consistent naming conventions:

• sqlglot_dialect for referring to the SQL dialect string used by the SQLGlot library.
• sql_dialect for the Splink dialect object.
• sql_dialect_str for Splink's string representation of the dialect.

One important test I did to make sure this works correctly is in dialects.py modify to:

class DuckDBDialect(SplinkDialect):
    _dialect_name_for_factory = "splink_ddb_alias"

    @property
    def sql_dialect_str(self):
        return "splink_ddb_alias"

    @property
    def sqlglot_dialect(self):
        return "duckdb"

and run

Click to expand

import splink.comparison_library as cl
from splink import DuckDBAPI, Linker, SettingsCreator, block_on, splink_datasets

db_api = DuckDBAPI()

df = splink_datasets.fake_1000

settings = SettingsCreator(
    link_type="dedupe_only",
    comparisons=[
        cl.ExactMatch("first_name"),
        cl.ExactMatch("surname"),
        cl.ExactMatch("dob"),
        cl.ExactMatch("city").configure(term_frequency_adjustments=True),
        cl.ExactMatch("email"),
    ],
    blocking_rules_to_generate_predictions=[
        block_on("first_name"),
        block_on("surname"),
    ],
    max_iterations=2,
)

linker = Linker(df, settings, db_api)

linker.training.estimate_probability_two_random_records_match(
    [block_on("first_name", "surname")], recall=0.7
)

linker.training.estimate_u_using_random_sampling(max_pairs=1e6)

linker.training.estimate_parameters_using_expectation_maximisation(block_on("dob"))

pairwise_predictions = linker.inference.predict(threshold_match_weight=-10)



settings2 = linker.misc.save_model_to_json()

linker2 = Linker(df, settings2, db_api)

linker2.inference.predict(threshold_match_weight=-10)

To validate that it works, and that splink_ddb_alias appears for sql_dialect in the json settings dictionary

@ADBond I've asked you for a review no urgency on this one!

[ADBond]

This is great!

Thanks for going through and fixing all of this, imagine it was fairly tedious to get this all aligned

[RobinL]

Thanks. Yes it was, ha ha! Though the type hinting/mypy work you did made it far easier