Snowflake row generator to insert random adhoc data in table - snowflake-cloud-data-platform

Question

I have a table t1 with one record.
name Reg address
david 12 a23 Carl Marx
Now I need to create 100 record using row generator and insert that into the table.
so that table t1 has 101 record now with 100 row of adhoc data in the table.
I can write a query like
Insert into t1 from
select randstr(5, random()) name,
uniform(1, 10, random(12)) reg,
randstr(20, random()) address
from table(generator(rowcount => 10));
However I have 40 table like t1 with 30 plus column.
So I need to use the datatype of table column to generate adhoc data.
Is there a better way to use the datatype of table to generate adhoc data

Answer 1

(Edited with usage instructions below) The Gods of SQL Purity may scorn me for this answer but I propose creating a Javascript UDF that takes the output of GET_DDL, parses the results and returns to a SQL statement, which you can then copy/paste and run.
Here is one such UDF which worked against the table I tried it on by select prepare_seed_stmt('my_tbl', get_ddl('table', 'my_tbl'));
Pay close attention to the expressions object to make sure it does what you want and since many datatypes are missing (like VARIANT, etc)
create or replace function prepare_seed_stmt(t string, ddl string)
RETURNS string
LANGUAGE JAVASCRIPT
AS
$$
const regex = /^\t(.*?) (.*?)(?: |,|\n)/gm;
const expressions = {
"string": "randstr(5, random()) ",
"number": "uniform(1, 10, random(12))",
"object": "object_construct(randstr(5, random()), uniform(1, 10, random(12)))",
"ts_ntz": "dateadd(second, uniform(-1e6, 1e6, random()), current_timestamp::timestamp_ntz)",
"array": "array_construct(uniform(1, 10, random(12))"
}
const mapping = {
"FLOAT": "number",
"NUMBER": "number",
"VARCHAR": "string",
"OBJECT": "object",
"ARRAY": "array",
"TIMESTAMP_NTZ": "ts_ntz",
}
let m;
let sql = `insert into ${T} from\nselect \n`;
while ((m = regex.exec(DDL)) !== null) {
if (m.index === regex.lastIndex) {
regex.lastIndex++;
}
sql += `${expressions[mapping[m[2].replace(/\(|\)|[0-9]/g, '')]]} ${m[1]},\n`
}
sql = sql.slice(0, -2)
sql += "\nfrom table(generator(rowcount => 10));"
return sql
$$
;
Usage
Say you have a table 'foo' and you want to generate a SQL insert into foo from select <for_each_column_an_expression_for_random_data> from table(generator(rowcount => 10))
FOO is defined as follows
select get_ddl('table', 'foo');
--returns:
create or replace TABLE FOO (
a NUMBER(38,0),
b VARCHAR(16777216),
c VARCHAR(16777216),
d VARCHAR(16777216),
e TIMESTAMP_NTZ(9),
f VARCHAR(16777216),
g FLOAT,
h OBJECT,
i OBJECT,
j OBJECT,
k OBJECT,
l VARCHAR(16777216),
m VARCHAR(16777216),
n OBJECT,
o OBJECT,
p VARCHAR(16777216),
q NUMBER(38,0),
r VARCHAR(16777216),
s NUMBER(38,0),
t TIMESTAMP_NTZ(9) DEFAULT CAST(CONVERT_TIMEZONE('UTC', CAST(CURRENT_TIMESTAMP() AS TIMESTAMP_TZ(9))) AS TIMESTAMP_NTZ(9))
);
Run select prepare_seed_sql('foo', get_ddl('table', 'foo')); to return the following, which can then be executed.
insert into foo from
select
uniform(1, 10, random(12)) a,
randstr(5, random()) b,
randstr(5, random()) c,
randstr(5, random()) d,
dateadd(second, uniform(-1e6, 1e6, random()), current_timestamp::timestamp_ntz) e,
randstr(5, random()) f,
uniform(1, 10, random(12)) g,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) h,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) i,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) j,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) k,
randstr(5, random()) l,
randstr(5, random()) m,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) n,
object_construct(randstr(5, random()), uniform(1, 10, random(12))) o,
randstr(5, random()) p,
uniform(1, 10, random(12)) q,
randstr(5, random()) r,
uniform(1, 10, random(12)) s,
dateadd(second, uniform(-1e6, 1e6, random()), current_timestamp::timestamp_ntz) t
from table(generator(rowcount => 10));