static rasqal_row* rasqal_join_rowsource_read_row(rasqal_rowsource* rowsource, void *user_data) { rasqal_join_rowsource_context* con; rasqal_row* row = NULL; rasqal_query *query = rowsource->query; con = (rasqal_join_rowsource_context*)user_data; if(con->failed || con->state == JS_FINISHED) return NULL; while(1) { rasqal_row *right_row; int bresult = 1; int compatible = 1; if(con->state == JS_START) { /* start / re-start left */ if(con->left_row) rasqal_free_row(con->left_row); con->left_row = rasqal_rowsource_read_row(con->left); #ifdef RASQAL_DEBUG RASQAL_DEBUG2("rowsource %p read left row : ", rowsource); if(con->left_row) rasqal_row_print(con->left_row, stderr); else fputs("NONE", stderr); fputs("\n", stderr); #endif con->state = JS_INIT_RIGHT; } if(con->state == JS_INIT_RIGHT) { /* start right */ if(!con->left_row) { con->state = JS_FINISHED; return NULL; } con->right_rows_joined_count = 0; rasqal_rowsource_reset(con->right); } right_row = rasqal_rowsource_read_row(con->right); #ifdef RASQAL_DEBUG RASQAL_DEBUG2("rowsource %p read right row : ", rowsource); if(right_row) rasqal_row_print(right_row, stderr); else fputs("NONE", stderr); fputs("\n", stderr); #endif if(!right_row && con->state == JS_READ_RIGHT) { /* right has finished */ /* restart left */ con->state = JS_START; /* if all right table returned no bindings, return left row */ if(!con->right_rows_joined_count) { /* otherwise return LEFT or RIGHT row only */ if(con->join_type == RASQAL_JOIN_TYPE_LEFT) { /* LEFT JOIN - add left row if expr fails or not compatible */ if(con->left_row) { con->right_rows_joined_count++; row = rasqal_join_rowsource_build_merged_row(rowsource, con, NULL); break; } } } /* restart left by continuing the loop */ continue; } /* state is always JS_READ_RIGHT at this point */ con->state = JS_READ_RIGHT; /* now may have both left and right rows so compute compatibility */ if(right_row) { compatible = rasqal_row_compatible_check(con->rc_map, con->left_row, right_row); RASQAL_DEBUG2("join rows compatible: %s\n", compatible ? "YES" : "NO"); } if(con->constant_join_condition >= 0) { /* Get constant join expression value */ bresult = con->constant_join_condition; } else if(con->expr) { /* Check join expression if present */ rasqal_literal *result; int error = 0; result = rasqal_expression_evaluate2(con->expr, query->eval_context, &error); #ifdef RASQAL_DEBUG RASQAL_DEBUG1("join expression result: "); if(error) fputs("type error", DEBUG_FH); else rasqal_literal_print(result, DEBUG_FH); fputc('\n', DEBUG_FH); #endif if(error) { bresult = 0; } else { error = 0; bresult = rasqal_literal_as_boolean(result, &error); #ifdef RASQAL_DEBUG if(error) RASQAL_DEBUG1("filter boolean expression returned error\n"); else RASQAL_DEBUG2("filter boolean expression result: %d\n", bresult); #endif rasqal_free_literal(result); } } if(con->join_type == RASQAL_JOIN_TYPE_NATURAL) { /* found a row if compatible and constraint matches */ if(compatible && bresult && right_row) { con->right_rows_joined_count++; /* consumes right_row */ row = rasqal_join_rowsource_build_merged_row(rowsource, con, right_row); break; } } else if(con->join_type == RASQAL_JOIN_TYPE_LEFT) { /* * { merge(mu1, mu2) | mu1 in Omega1 and mu2 in Omega2, and mu1 * and mu2 are compatible and expr(merge(mu1, mu2)) is true } */ if(compatible && bresult) { con->right_rows_joined_count++; /* No constraint OR constraint & compatible so return merged row */ /* Compute row only now it is known to be needed (consumes right_row) */ row = rasqal_join_rowsource_build_merged_row(rowsource, con, right_row); break; } #if 0 /* * { mu1 | mu1 in Omega1 and mu2 in Omega2, and mu1 and mu2 are * not compatible } */ if(!compatible) { /* otherwise return LEFT or RIGHT row only */ if(con->join_type == RASQAL_JOIN_TYPE_LEFT) { /* LEFT JOIN - add left row if expr fails or not compatible */ if(con->left_row) { con->right_rows_joined_count++; row = rasqal_join_rowsource_build_merged_row(rowsource, con, NULL); if(right_row) rasqal_free_row(right_row); break; } } } #endif /* * { mu1 | mu1 in Omega1 and mu2 in Omega2, and mu1 and mu2 are * compatible and for all mu2, expr(merge(mu1, mu2)) is false } */ /* The above is handled using check for * !con->right_rows_joined_count earlier, to generate a row * once. */ } /* end if LEFT JOIN */ if(right_row) rasqal_free_row(right_row); } /* end while */ if(row) { rasqal_row_set_rowsource(row, rowsource); row->offset = con->offset++; rasqal_row_bind_variables(row, rowsource->query->vars_table); } return row; }
static rasqal_row* rasqal_aggregation_rowsource_read_row(rasqal_rowsource* rowsource, void *user_data) { rasqal_aggregation_rowsource_context* con; rasqal_row* row; int error = 0; con = (rasqal_aggregation_rowsource_context*)user_data; if(con->finished) return NULL; /* Iterate over input rows until last row seen or group done */ while(1) { error = 0; if(con->saved_row) row = con->saved_row; else row = rasqal_rowsource_read_row(con->rowsource); if(!row) { /* End of input - calculate last aggregation result */ con->finished = 1; break; } if(con->last_group_id != row->group_id) { int i; if(!con->saved_row && con->last_group_id >= 0) { /* Existing aggregation is done - return result */ /* save current row for next time this function is called */ con->saved_row = row; row = NULL; #ifdef RASQAL_DEBUG RASQAL_DEBUG2("Aggregation ending group %d", con->last_group_id); fputc('\n', DEBUG_FH); #endif /* Empty distinct maps */ for(i = 0; i < con->expr_count; i++) { rasqal_agg_expr_data* expr_data = &con->expr_data[i]; if(expr_data->map) { rasqal_free_map(expr_data->map); expr_data->map = NULL; } } break; } /* reference is now in 'row' variable */ con->saved_row = NULL; #ifdef RASQAL_DEBUG RASQAL_DEBUG2("Aggregation starting group %d", row->group_id); fputc('\n', DEBUG_FH); #endif /* next time this function is called we continue here */ for(i = 0; i < con->expr_count; i++) { rasqal_agg_expr_data* expr_data = &con->expr_data[i]; if(!expr_data->agg_user_data) { /* init once */ expr_data->agg_user_data = rasqal_builtin_agg_expression_execute_init(rowsource->world, expr_data->expr); if(!expr_data->agg_user_data) { error = 1; break; } } /* Init map for each group */ if(expr_data->expr->flags & RASQAL_EXPR_FLAG_DISTINCT) { expr_data->map = rasqal_new_literal_sequence_sort_map(1 /* is_distinct */, 0 /* compare_flags */); if(!expr_data->map) { error = 1; break; } } } if(error) break; con->last_group_id = row->group_id; } /* end if handling change of group ID */ /* Bind the values in the input row to the variables in the table */ rasqal_row_bind_variables(row, rowsource->query->vars_table); /* Evaluate the expressions giving a sequence of literals to * run the aggregation step over. */ if(1) { int i; if(!con->step_count) { /* copy first value row from input rowsource */ for(i = 0; i < con->input_values_count; i++) { rasqal_literal* value; value = rasqal_new_literal_from_literal(row->values[i]); raptor_sequence_set_at(con->input_values, i, value); } } con->step_count++; for(i = 0; i < con->expr_count; i++) { rasqal_agg_expr_data* expr_data = &con->expr_data[i]; raptor_sequence* seq; /* SPARQL Aggregation uses ListEvalE() to evaluate - ignoring * errors and filtering out expressions that fail */ seq = rasqal_expression_sequence_evaluate(rowsource->query, expr_data->exprs_seq, /* ignore_errors */ 1, &error); if(error) continue; if(expr_data->map) { if(rasqal_literal_sequence_sort_map_add_literal_sequence(expr_data->map, seq)) { /* duplicate found * * The above function just freed seq so no data is lost */ continue; } } #ifdef RASQAL_DEBUG RASQAL_DEBUG1("Aggregation step over literals: "); raptor_sequence_print(seq, DEBUG_FH); fputc('\n', DEBUG_FH); #endif error = rasqal_builtin_agg_expression_execute_step(expr_data->agg_user_data, seq); /* when DISTINCTing, seq remains owned by the map * otherwise seq is local and must be freed */ if(!expr_data->map) raptor_free_sequence(seq); if(error) break; } } rasqal_free_row(row); row = NULL; if(error) break; } /* end while reading rows */ if(error) { /* Discard row on error */ if(row) { rasqal_free_row(row); row = NULL; } } else if (con->last_group_id >= 0) { int offset = 0; int i; /* Generate result row and reset for next group */ row = rasqal_new_row(rowsource); /* Copy scalar results through */ for(i = 0; i < con->input_values_count; i++) { rasqal_literal* result; /* Reset: get and delete any stored input rowsource literal */ result = (rasqal_literal*)raptor_sequence_delete_at(con->input_values, i); rasqal_row_set_value_at(row, offset, result); rasqal_free_literal(result); offset++; } /* Set aggregate results */ for(i = 0; i < con->expr_count; i++) { rasqal_literal* result; rasqal_agg_expr_data* expr_data = &con->expr_data[i]; rasqal_variable* v; /* Calculate the result because the input ended or a new group started */ result = rasqal_builtin_agg_expression_execute_result(expr_data->agg_user_data); #ifdef RASQAL_DEBUG RASQAL_DEBUG1("Aggregation ending group with result: "); if(result) rasqal_literal_print(result, DEBUG_FH); else fputs("NULL", DEBUG_FH); fputc('\n', DEBUG_FH); #endif v = rasqal_rowsource_get_variable_by_offset(rowsource, offset); result = rasqal_new_literal_from_literal(result); /* it is OK to bind to NULL */ rasqal_variable_set_value(v, result); rasqal_row_set_value_at(row, offset, result); if(result) rasqal_free_literal(result); offset++; if(rasqal_builtin_agg_expression_execute_reset(expr_data->agg_user_data)) { rasqal_free_row(row); row = NULL; break; } } con->step_count = 0; if(row) row->offset = con->offset++; } return row; }
static int rasqal_groupby_rowsource_process(rasqal_rowsource* rowsource, rasqal_groupby_rowsource_context* con) { /* already processed */ if(con->processed) return 0; con->processed = 1; /* Empty expression list - no need to read rows */ if(!con->expr_seq || !con->expr_seq_size) { con->group_id++; return 0; } con->tree = raptor_new_avltree(rasqal_rowsource_groupby_literal_sequence_compare, (raptor_data_free_handler)rasqal_free_groupby_tree_node, /* flags */ 0); if(!con->tree) return 1; raptor_avltree_set_print_handler(con->tree, rasqal_rowsource_groupby_tree_print_node); while(1) { rasqal_row* row; row = rasqal_rowsource_read_row(con->rowsource); if(!row) break; rasqal_row_bind_variables(row, rowsource->query->vars_table); if(con->expr_seq) { raptor_sequence* literal_seq; rasqal_groupby_tree_node key; rasqal_groupby_tree_node* node; literal_seq = rasqal_expression_sequence_evaluate(rowsource->query, con->expr_seq, /* ignore_errors */ 0, /* literal_seq */ NULL, /* error_p */ NULL); if(!literal_seq) { /* FIXME - what to do on errors? */ continue; } memset(&key, '\0', sizeof(key)); key.con = con; key.literals = literal_seq; node = (rasqal_groupby_tree_node*)raptor_avltree_search(con->tree, &key); if(!node) { /* New Group */ node = (rasqal_groupby_tree_node*)RASQAL_CALLOC(rasqal_groupby_tree_node, sizeof(*node), 1); if(!node) { raptor_free_sequence(literal_seq); return 1; } node->con = con; node->group_id = ++con->group_id; /* node now owns literal_seq */ node->literals = literal_seq; #ifdef HAVE_RAPTOR2_API node->rows = raptor_new_sequence((raptor_data_free_handler)rasqal_free_row, (raptor_data_print_handler)rasqal_row_print); #else node->rows = raptor_new_sequence((raptor_sequence_free_handler*)rasqal_free_row, (raptor_sequence_print_handler*)rasqal_row_print); #endif if(!node->rows) { rasqal_free_groupby_tree_node(node); return 1; } /* after this, node is owned by con->tree */ raptor_avltree_add(con->tree, node); } else raptor_free_sequence(literal_seq); row->group_id = node->group_id; /* after this, node owns the row */ raptor_sequence_push(node->rows, row); } } #ifdef RASQAL_DEBUG if(con->tree) { fputs("Grouping ", DEBUG_FH); raptor_avltree_print(con->tree, DEBUG_FH); fputs("\n", DEBUG_FH); } #endif con->group_iterator = raptor_new_avltree_iterator(con->tree, NULL, NULL, 1); con->group_row_index = 0; con->offset = 0; return 0; }