使用Lambda并发流翻车事件——parallelStream().filter()过滤出来的个数少了

场景：

处理一个list集合：List<CharVar> dpsCharVars，过滤不同的映射集合，并进行下一步处理。

        JSONObject onehot_mappings = new JSONObject();
        JSONObject seq_mappings = new JSONObject();
        JSONObject woe_mappings = new JSONObject();
        
        List<CharVar> oneHotList = dpsCharVars.parallelStream()
                .filter(CharVar::iisOneHotMapping)
                .forEach(dpsChar -> onehot_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon(), List.class)));
        log.info("====================onehot映射大小{} ：{}==========================", onehot_mappings.size(), onehot_mappings);
        //顺序替换的变量 的映射规则  {"col1":{"a":1}, "col2":{"b":2} }
        List<CharVar> orderList = dpsCharVars.parallelStream()
                .filter(CharVar::iisOrderMapping)
                .forEach(dpsChar -> seq_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon())));
        log.info("====================顺序映射大小{} ：{}==========================", seq_mappings.size(), seq_mappings);
        //按woe替换的变量 的映射规则 {"col3":{"a":1}, "col4":{"b":2} }
        List<CharVar> woeList = dpsCharVars.parallelStream()
                .filter(CharVar::iisWoeMapping)
                .forEac(dpsChar -> woe_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon())));
        log.info("====================woe映射大小{} ：{}==========================", woe_mappings.size(), woe_mappings);

结果：

运行多次，不是onehot_mappings少一两个值，就是seq_mappings少一两个值。

原因：

这是因为使用了并发流parallelStream().

解决办法：

方法一：

将过滤出来的流先转成list，再进行计算

        List<CharVar> oneHotList = dpsCharVars.parallelStream()
                .filter(CharVar::iisOneHotMapping)
                .collect(Collectors.toList());
        oneHotList.forEach(dpsChar -> onehot_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon(), List.class)));
        log.info("====================onehot映射大小{} ：{}==========================", onehot_mappings.size(), onehot_mappings);
        //顺序替换的变量 的映射规则  {"col1":{"a":1}, "col2":{"b":2} }
        List<CharVar> orderList = dpsCharVars.parallelStream()
                .filter(CharVar::iisOrderMapping)
                .collect(Collectors.toList());
        orderList.forEach(dpsChar -> seq_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon())));
        log.info("====================顺序映射大小{} ：{}==========================", seq_mappings.size(), seq_mappings);
        //按woe替换的变量 的映射规则 {"col3":{"a":1}, "col4":{"b":2} }
        List<CharVar> woeList = dpsCharVars.parallelStream()
                .filter(CharVar::iisWoeMapping)
                .collect(Collectors.toList());
        woeList.forEach(dpsChar -> woe_mappings.put(dpsChar.getEnName(), JSONObject.parseObject(dpsChar.getVarCon())));
        log.info("====================woe映射大小{} ：{}==========================", woe_mappings.size(), woe_mappings);

方法二：推荐

        dpsCharVars.forEach(var -> {
            if (var.iisOneHotMapping()) {
                onehot_mappings.put(var.getEnName(), JSONObject.parseObject(var.getVarCon(), List.class));
            } else if (var.iisOrderMapping()) {
                seq_mappings.put(var.getEnName(), JSONObject.parseObject(var.getVarCon()));
            } else if (var.iisWoeMapping()) {
                woe_mappings.put(var.getEnName(), JSONObject.parseObject(var.getVarCon()));
            } else {
                log.info("未知映射类型：{}", var);
            }
        });
        log.info("====================onehot映射大小{} ：{}==========================", onehot_mappings.size(), onehot_mappings);
        log.info("====================顺序映射大小{} ：{}==========================", seq_mappings.size(), seq_mappings);
        log.info("====================woe映射大小{} ：{}==========================", woe_mappings.size(), woe_mappings);