Examples with DfDelimiterDataFirstLineInfo org.dbflute.logic.replaceschema.loaddata.delimiter.line.DfDelimiterDataFirstLineInfo used on opensource projects

Example 1 with DfDelimiterDataFirstLineInfo

use of org.dbflute.logic.replaceschema.loaddata.delimiter.line.DfDelimiterDataFirstLineInfo in project dbflute-core by dbflute.

the class DfDelimiterDataWriterImpl method doWriteData.

// -----------------------------------------------------
// Write Data
// ----------
protected void doWriteData(DfDelimiterDataResultInfo resultInfo, boolean forcedlySuppressBatch, int offsetRowCount) throws IOException {
    final String dataDirectory = Srl.substringLastFront(_filePath, "/");
    final LoggingInsertType loggingInsertType = getLoggingInsertType(dataDirectory);
    final String tableDbName = extractTableDbName();
    final Map<String, DfColumnMeta> columnMetaMap = getColumnMetaMap(tableDbName);
    if (columnMetaMap.isEmpty()) {
        throwTableNotFoundException(_filePath, tableDbName);
    }
    // process before handling table
    beforeHandlingTable(tableDbName, columnMetaMap);
    // fixedly
    final String lineSeparatorInValue = "\n";
    final File dataFile = new File(_filePath);
    final boolean canBatchUpdate = canBatchUpdate(forcedlySuppressBatch, dataDirectory);
    final StringBuilder lineStringSb = new StringBuilder();
    final StringBuilder preContinuedSb = new StringBuilder();
    final List<String> columnNameList = new ArrayList<String>();
    final List<String> columnValueList = new ArrayList<String>();
    List<String> valueListSnapshot = null;
    // not line on file, as registered record
    int rowNumber = 0;
    String executedSql = null;
    // may committed per limit size, for skip in retry
    int committedRowCount = 0;
    FileInputStream fis = null;
    InputStreamReader ir = null;
    BufferedReader br = null;
    Connection conn = null;
    PreparedStatement ps = null;
    try {
        fis = new FileInputStream(dataFile);
        ir = new InputStreamReader(fis, _encoding);
        br = new BufferedReader(ir);
        DfDelimiterDataFirstLineInfo firstLineInfo = null;
        int loopIndex = -1;
        // current registered size to prepared statement
        int addedBatchSize = 0;
        while (true) {
            ++loopIndex;
            {
                final String readLine = br.readLine();
                if (readLine == null) {
                    break;
                }
                clearAppend(lineStringSb, readLine);
            }
            // - - - - - - - - - -/
            if (loopIndex == 0) {
                firstLineInfo = analyzeFirstLine(lineStringSb.toString(), _delimiter);
                setupColumnNameList(columnNameList, dataDirectory, dataFile, tableDbName, firstLineInfo, columnMetaMap);
                continue;
            }
            // /- - - - - - - - - - - - - - -
            // analyze values in line strings
            // - - - - - - - - - -/
            // might be clear-appended
            filterLineStringIfNeeds(lineStringSb);
            {
                if (preContinuedSb.length() > 0) {
                    // done performance tuning, suppress incremental strings from many line separators by jflute (2018/03/02)
                    // it needs to change lineString, preContinueString to StringBuilder type...
                    // lineString = preContinueString + "\n" + lineString; (2021/01/21)
                    // and insert has array-copy so may not be fast
                    // lineStringSb.insert(0, "\n").insert(0, preContinuedSb); (2021/01/21)
                    // used only here so changing is no problem
                    preContinuedSb.append(lineSeparatorInValue).append(lineStringSb);
                    clearAppend(lineStringSb, preContinuedSb);
                }
                final DfDelimiterDataValueLineInfo valueLineInfo = analyzeValueLine(lineStringSb.toString(), _delimiter);
                // empty string resolved later
                final List<String> extractedList = valueLineInfo.getValueList();
                if (valueLineInfo.isContinueNextLine()) {
                    clearAppend(preContinuedSb, extractedList.remove(extractedList.size() - 1));
                    columnValueList.addAll(extractedList);
                    // keeping valueList that has previous values
                    continue;
                }
                columnValueList.addAll(extractedList);
            }
            // - - - - - - - - - -/
            if (isDifferentColumnValueCount(firstLineInfo, columnValueList)) {
                handleDifferentColumnValueCount(resultInfo, dataDirectory, tableDbName, firstLineInfo, columnValueList);
                // clear temporary variables
                clear(preContinuedSb);
                columnValueList.clear();
                valueListSnapshot = null;
                continue;
            }
            // *valid record is prepared here
            ++rowNumber;
            valueListSnapshot = columnValueList;
            if (rowNumber <= offsetRowCount) {
                // basically only when retry
                // clear temporary variables
                clear(preContinuedSb);
                columnValueList.clear();
                valueListSnapshot = null;
                // e.g. 1 ~ 100000 rows if 100000 already committed
                continue;
            }
            // /- - - - - - - - - - - - - - - -
            // process registration to database
            // - - - - - - - - - -/
            final DfDelimiterDataWriteSqlBuilder sqlBuilder = createSqlBuilder(resultInfo, tableDbName, columnMetaMap, columnNameList, columnValueList);
            if (conn == null) {
                conn = _dataSource.getConnection();
            }
            if (ps == null) {
                // for performance (suppress implicit transaction per SQL)
                beginTransaction(conn);
                executedSql = sqlBuilder.buildSql();
                ps = prepareStatement(conn, executedSql);
            }
            final Map<String, Object> columnValueMap = sqlBuilder.setupParameter();
            final Set<String> sysdateColumnSet = sqlBuilder.getSysdateColumnSet();
            resolveRelativeDate(dataDirectory, tableDbName, columnValueMap, columnMetaMap, sysdateColumnSet, rowNumber);
            handleLoggingInsert(tableDbName, columnValueMap, loggingInsertType, rowNumber);
            int bindCount = 1;
            for (Entry<String, Object> entry : columnValueMap.entrySet()) {
                final String columnName = entry.getKey();
                final Object obj = entry.getValue();
                // - - - - - - - - - -/
                if (processNull(dataDirectory, tableDbName, columnName, obj, ps, bindCount, columnMetaMap, rowNumber)) {
                    bindCount++;
                    continue;
                }
                // It registers the value to statement by the type.
                if (processNotNullNotString(dataDirectory, tableDbName, columnName, obj, conn, ps, bindCount, columnMetaMap, rowNumber)) {
                    bindCount++;
                    continue;
                }
                // /- - - - - - - - - - - - - - - - - -
                // process NotNull and StringExpression
                // - - - - - - - - - -/
                final String value = (String) obj;
                processNotNullString(dataDirectory, dataFile, tableDbName, columnName, value, conn, ps, bindCount, columnMetaMap, rowNumber);
                bindCount++;
            }
            if (canBatchUpdate) {
                // mainly here
                ps.addBatch();
            } else {
                ps.execute();
            }
            ++addedBatchSize;
            if (isBatchLimit(dataDirectory, addedBatchSize)) {
                // transaction scope
                if (canBatchUpdate) {
                    // mainly here
                    // this is supported in only delimiter data writer because delimiter data can treat large data
                    // (actually needed, GC overhead limit exceeded when 1000000 records to MySQL, 2021/01/20)
                    // to avoid OutOfMemory
                    ps.executeBatch();
                }
                commitTransaction(conn);
                committedRowCount = committedRowCount + addedBatchSize;
                addedBatchSize = 0;
                close(ps);
                ps = null;
            }
            // *one record is finished here
            // clear temporary variables
            // if an exception occurs from execute() or addBatch(),
            // this valueList is to be information for debug
            clear(preContinuedSb);
            columnValueList.clear();
        // keep here for retry
        // valueListSnapshot = null;
        }
        if (ps != null && addedBatchSize > 0) {
            if (canBatchUpdate) {
                // mainly here
                ps.executeBatch();
            }
            commitTransaction(conn);
            committedRowCount = committedRowCount + addedBatchSize;
        }
        noticeLoadedRowSize(tableDbName, rowNumber);
        resultInfo.registerLoadedMeta(dataDirectory, _filePath, rowNumber);
        checkImplicitClassification(dataFile, tableDbName, columnNameList);
    } catch (SQLException e) {
        // request retry if it needs (e.g. execution exception of batch insert)
        // the snapshot is used only when retry failure basically
        final DfJDBCException wrapped = DfJDBCException.voice(e);
        final String msg = buildFailureMessage(_filePath, tableDbName, executedSql, columnValueList, wrapped);
        throw new DfDelimiterDataRegistrationFailureException(msg, wrapped.getNextException()).retryIfNeeds(createRetryResource(canBatchUpdate, committedRowCount)).snapshotRow(createRowSnapshot(columnNameList, valueListSnapshot, rowNumber));
    } catch (RuntimeException e) {
        // unneeded snapshot at this side but just in case (or changing determination future)
        final String msg = buildFailureMessage(_filePath, tableDbName, executedSql, columnValueList, null);
        throw new DfDelimiterDataRegistrationFailureException(msg, e).snapshotRow(createRowSnapshot(columnNameList, valueListSnapshot, rowNumber));
    } finally {
        closeStream(fis, ir, br);
        try {
            rollbackTransaction(conn);
        } catch (SQLException continued) {
            _log.info("Failed to rollback the delimiter data transaction.", continued);
        }
        close(ps);
        close(conn);
        // process after (finally) handling table
        finallyHandlingTable(tableDbName, columnMetaMap);
    }
}