void from_recordset_stream(const std::string &reliable_storage_local_root,
const std::string &reliable_storage_remote_root,
Input_Stream &input, Output_Stream &output,
const std::vector<rel::rlang::token> &tokens) {
/* Get the headers. */
record headings(input.parse_headings());
// Get the resource id field id, this is the only field that we're interested
// in, the rest are just placeholders.
size_t resource_id_field_id = headings.mandatory_find_field(
NP1_REL_DISTRIBUTED_RESOURCE_ID_FIELD_NAME);
// Create the target recordset.
std::string target_recordset_name(rel::rlang::compiler::eval_to_string_only(tokens));
np1::io::reliable_storage::id target_recordset_resource_id(target_recordset_name);
np1::io::reliable_storage rs(reliable_storage_local_root, reliable_storage_remote_root);
np1::io::reliable_storage::stream target_recordset_stream(rs);
NP1_ASSERT(rs.create_wo(target_recordset_resource_id, target_recordset_stream),
"Unable to create recordset " + target_recordset_name);
// Read all the input recordset chunks ids and write them to the recordset file.
input.parse_records(
shallow_copy_chunk_record_callback(
rs, target_recordset_stream, resource_id_field_id, headings.ref()));
NP1_ASSERT(target_recordset_stream.close(),
"Unable to close target recordset stream");
}
void from_data_stream(const std::string &reliable_storage_local_root,
const std::string &reliable_storage_remote_root,
Input_Stream &input, Output_Stream &output,
const std::vector<rel::rlang::token> &tokens) {
/* Get the headers. */
record headings(input.parse_headings());
// Interpret the arguments.
std::vector<std::pair<std::string, rlang::dt::data_type> > args = rel::rlang::compiler::eval_to_strings(tokens);
NP1_ASSERT((args.size() > 0) && (args.size() <= 2), "Incorrect number of arguments to rel.recordset.create");
tokens[0].assert(rlang::dt::data_type::TYPE_STRING == args[0].second,
"First argument to rel.recordset.create is not a string");
std::string target_recordset_name = args[0].first;
uint64_t approx_max_chunk_size = DEFAULT_APPROX_MAX_CHUNK_SIZE_BYTES;
if (args.size() > 1) {
tokens[0].assert((rlang::dt::data_type::TYPE_INT == args[1].second)
|| (rlang::dt::data_type::TYPE_UINT == args[1].second),
"Second argument to rel.recordset.create is not an integer");
approx_max_chunk_size = str::dec_to_int64(args[1].first);
}
// Get the stream that will hold the recordset.
np1::io::reliable_storage::id target_recordset_resource_id(target_recordset_name);
np1::io::reliable_storage rs(reliable_storage_local_root, reliable_storage_remote_root);
np1::io::reliable_storage::stream target_recordset_stream(rs);
NP1_ASSERT(rs.create_wo(target_recordset_resource_id, target_recordset_stream),
"Unable to create recordset " + target_recordset_name
+ " in reliable storage '" + reliable_storage_local_root + "'");
np1::io::reliable_storage::stream current_target_chunk_stream(rs);
buffered_reliable_storage_stream_type buffered_current_target_chunk_stream(
current_target_chunk_stream);
mandatory_reliable_storage_stream_type mandatory_current_target_chunk_stream(
buffered_current_target_chunk_stream);
np1::io::reliable_storage::id current_target_chunk_id;
uint64_t current_target_chunk_size = 0;
// Read all the input data and redistribute it into recordset chunks.
input.parse_records(
chunk_data_record_callback(
rs, target_recordset_stream, current_target_chunk_id,
current_target_chunk_stream, mandatory_current_target_chunk_stream,
current_target_chunk_size, approx_max_chunk_size, headings.ref()));
// Close everything.
mandatory_current_target_chunk_stream.close();
NP1_ASSERT(target_recordset_stream.close(),
"Unable to close target recordset stream");
}
void operator()(Input_Stream &input, Output_Stream &output,
const rstd::vector<rel::rlang::token> &tokens) {
NP1_ASSERT(tokens.size() == 0, "rel.record_count accepts no arguments");
// Read & discard the headings.
input.parse_headings();
uint64_t number_records = 0;
input.parse_records(record_counter_callback(number_records));
output.write(str::to_dec_str(number_records).c_str());
}
static void run(Input_Stream &input, Final_Output_Stream &output,
const rstd::vector<rel::rlang::token> &tokens) {
/* Get the headers and the interesting fields out of them. */
rel::record input_headings(input.parse_headings());
size_t file_name_field_id =
input_headings.mandatory_find_heading(NP1_META_PARALLEL_EXPLICIT_MAPPING_HEADING_FILE_NAME);
size_t host_name_field_id =
input_headings.mandatory_find_heading(NP1_META_PARALLEL_EXPLICIT_MAPPING_HEADING_HOST_NAME);
// The object that manages all the processes.
process_pool_map_type process_pool_map;
bool output_headings_written = false;
// Translate all the input records into command lines and put them in the pool ready to execute.
input.parse_records(input_record_callback(file_name_field_id, host_name_field_id, tokens,
process_pool_map, output, output_headings_written));
// Run all the processes in the pool.
process_pool_map.wait_all();
}
void operator()(Input_Stream &input, Output_Stream &output,
const std::vector<rel::rlang::token> &tokens,
sort_type_type sort_type,
sort_order_type sort_order) {
NP1_ASSERT(tokens.size() > 0, "Unexpected empty stream operator argument list");
// Read the first line of input, we need it to add meaning to the arguments.
record headings(input.parse_headings());
std::vector<std::string> arg_headings;
rlang::compiler::compile_heading_name_list(
tokens, headings.ref(), arg_headings);
// Create the compare specs.
detail::compare_specs comp_specs(headings, arg_headings);
// Write out the headings then do the actual sorting.
headings.write(output);
less_than lt(comp_specs);
greater_than gt(comp_specs);
switch (sort_type) {
case TYPE_MERGE_SORT:
switch (sort_order) {
case ORDER_ASCENDING:
sort<detail::merge_sort>(input, output, lt);
break;
case ORDER_DESCENDING:
sort<detail::merge_sort>(input, output, gt);
break;
}
break;
case TYPE_QUICK_SORT:
//TODO: why the dickens isn't this quick sort and why is quick sort broken?
switch (sort_order) {
case ORDER_ASCENDING:
sort<detail::merge_sort>(input, output, lt);
break;
case ORDER_DESCENDING:
sort<detail::merge_sort>(input, output, gt);
break;
}
break;
}
}
void operator()(const std::string &reliable_storage_local_root,
const std::string &reliable_storage_remote_root,
const std::string &listen_endpoint,
Input_Stream &input, Output_Stream &output,
const std::vector<rel::rlang::token> &tokens) {
log_info("Reading headers and compiling expression against headers.");
/* Get the headers. */
record headings(input.parse_headings());
// Compile, just to check that the headings & tokens are likely to work
// when we distribute.
record empty_headings;
rlang::vm vm = rlang::compiler::compile_single_expression(
tokens, headings.ref(), empty_headings.ref());
// Check that the expression is actually a boolean expression.
NP1_ASSERT(vm.return_type() == rlang::dt::TYPE_BOOL,
"Expression is not a boolean expression");
// Do the distribution.
distributed::distribute(log_id(), headings, headings, "rel.where", reliable_storage_local_root,
reliable_storage_remote_root, listen_endpoint, input, output, tokens);
}
