possible memory leak through ref cursor (analysis)

[c-bik]

@shamis and @acautin,

My erlang level attemps only confirms your observations so far. The test module mec_oc here is slightly enhanced version on @acautin 's work.

mec_oc:test_ref_cursor(Conn, TestSql, 1000, BindVars, BindValues).
{1000, 0.01%} {900, 0.02%} {800, 0.03%} {700, 0.03%} {600, 0.04%} {500, 0.04%} {400, 0.05%}
{300, 0.05%} {200, 0.06%} {100, 0.06%}

I think this module is already closest to dpi (as far erlang can go). Tried different combinations without any improvements so far!

Next step is to implement this with C (taking the NIF part out of the way to get losest to dpi APIs) and perform similar tests.

[c-bik]

{1000, 0.01%} {900, 0.02%} {800, 0.02%} {700, 0.02%} {600, 0.02%} {500, 0.02%} {400, 0.03%}
{300, 0.03%} {200, 0.03%} {100, 0.03%} {end, 0.03%}

[c-bik]

{1000, 0.01%} {900, 0.01%} {800, 0.02%} {700, 0.02%} {600, 0.02%} {500, 0.02%} {400, 0.02%}
{300, 0.03%} {200, 0.03%} {100, 0.03%} {end, 0.04%}

[c-bik]

> mec_oc:test_ref_cursor(Conn, TestSql, 1000, BindVars, BindValues).
Resources #{connection => 1,context => 1,data => 0,datapointer => 0,
            statement => 0,variable => 0}
{1000, 0.01%} {900, 0.01%} {800, 0.01%} {700, 0.01%} {600, 0.02%} {500, 0.02%} {400, 0.02%}
 {300, 0.02%} {200, 0.02%} {100, 0.02%} {end, 0.02%}
Resources #{connection => 1,context => 1,data => 0,datapointer => 0,
            statement => 0,variable => 0}