added postgres type-descriptor

[PrasannaIITM]

Link to issue: #142

[eyurtsev]

@PrasannaIITM Contribution very much appreciated!! Code is looking great! I left a question about matching postgres syntax

[eyurtsev]

The schema looks like it's something between typescript and postgres. I don't know how much this stuff matters since it's really hard to know what the LLM will do.

I was originally thinking about getting the postgres type description to output either the CREATE TABLE or a markdown table that shows the column/types/description of the table.

What are your thoughts on this?

CREATE TABLE sample_table (
    id SERIAL PRIMARY KEY,
    text_field TEXT,
    text_array_field TEXT[],
    float_field FLOAT,
    integer_field INTEGER,
    integer_array_field INTEGER[],
    embedded_json_field JSONB,
    color_enum_field COLOR_ENUM
);

CREATE TYPE color_enum AS ENUM ('RED', 'GREEN', 'BLUE');

or else the output from running:

SELECT column_name, data_type FROM information_schema.columns WHERE table_name = '...';   

something of this sort:

 column_name |     data_type
-------------+-------------------
 id          | integer
 text_field  | text
 text_array_field | text[]
 float_field | double precision
 integer_field | integer
 integer_array_field | integer[]
 embedded_json_field | jsonb
 color_enum_field | color_enum

[PrasannaIITM]

@eyurtsev You are right. I was still determining the exact syntax; the markdown syntax table looks correct. I will update my PR.

[PrasannaIITM]

One issue with the above approach is displaying nested objects. What would be the correct way to display nested objects in the markdown table format?

[eyurtsev]

Sorry for the delay -- was away from my computer yesterday!

I think this type descriptor best coupled to the CSV encoder.

The CSV encoder will likely rely on a JSON field for encoding nested objects, and that can be represented using the JSON type in postgres https://www.postgresql.org/docs/current/datatype-json.html

---

In either representation we choose (i.e., either the markdown table or the CREATE TABLE representation), we'll need to include the description of the column. An extra column for description would work for markdown and SQL style comments for the CREATE TABLE representation.