使用JSON.NET解析1GB json文件的问题

时间:2015-06-12 22:04:20

标签: json parsing stream json.net large-object-heap

我已经获得了一个应用程序,其中输入已从50K位置记录扩展到110万个位置记录。 这导致了严重的问题,因为整个文件先前被反序列化为单个对象。 对于具有110万条记录的生产类文件,对象的大小约为1GB。 由于大对象GC问题,我希望将反序列化对象保持在85K以下。

我试图一次解析一个位置对象并对其进行反序列化,以便控制对象的数量 得到反序列化,然后控制对象的大小。我使用Json.Net库来做到这一点。

以下是我作为流进入我的应用程序的JSON文件示例。

    {
    "Locations": [{
        "LocationId": "",
        "ParentLocationId": "",
        "DisplayFlag": "Y",
        "DisplayOptions": "",
        "DisplayName": "",
        "Address": "",
        "SecondaryAddress": "",
        "City": "",
        "State": "",
        "PostalCode": "",
        "Country": "",
        "Latitude": 40.59485,
        "Longitude": -73.96174,
        "LatLonQuality": 99,
        "BusinessLogoUrl": "",
        "BusinessUrl": "",
        "DisplayText": "",
        "PhoneNumber": "",
        "VenueGroup": 7,
        "VenueType": 0,
        "SubVenue": 0,
        "IndoorFlag": "",
        "OperatorDefined": "",
        "AccessPoints": [{
            "AccessPointId": "",
            "MACAddress": "",
            "DisplayFlag": "",
            "DisplayOptions": "",
            "Latitude": 40.59485,
            "Longitude": -73.96174,
            "Status": "Up",
            "OperatorDefined": "",
            "RoamingGroups": [{
                "GroupName": ""
            },
            {
                "GroupName": ""
            }],
            "Radios": [{
                "RadioId": "",
                "RadioFrequency": "",
                "RadioProtocols": [{
                    "Protocol": ""
                }],
                "WifiConnections": [{
                    "BSSID": "",
                    "ServiceSets": [{
                        "SSID": "",
                        "SSID_Broadcasted": ""
                    }]
                }]
            }]
        }]
    },
    {
        "LocationId": "",
        "ParentLocationId": "",
        "DisplayFlag": "Y",
        "DisplayOptions": "",
        "DisplayName": "",
        "Address": "",
        "SecondaryAddress": "",
        "City": "",
        "State": "",
        "PostalCode": "",
        "Country": "",
        "Latitude": 40.59485,
        "Longitude": -73.96174,
        "LatLonQuality": 99,
        "BusinessLogoUrl": "",
        "BusinessUrl": "",
        "DisplayText": "",
        "PhoneNumber": "",
        "VenueGroup": 7,
        "VenueType": 0,
        "SubVenue": 0,
        "IndoorFlag": "",
        "OperatorDefined": "",
        "AccessPoints": [{
            "AccessPointId": "",
            "MACAddress": "",
            "DisplayFlag": "",
            "DisplayOptions": "",
            "Latitude": 40.59485,
            "Longitude": -73.96174,
            "Status": "Up",
            "OperatorDefined": "",
            "RoamingGroups": [{
                "GroupName": ""
            },
            {
                "GroupName": ""
            }],
            "Radios": [{
                "RadioId": "",
                "RadioFrequency": "",
                "RadioProtocols": [{
                    "Protocol": ""
                }],
                "WifiConnections": [{
                    "BSSID": "",
                    "ServiceSets": [{
                        "SSID": "",
                        "SSID_Broadcasted": ""
                    }]
                }]
            }]
        }]
    }]
}

我需要能够拉出各个Location对象,以便我查看以下内容

    {
    "LocationId": "",
    "ParentLocationId": "",
    "DisplayFlag": "Y",
    "DisplayOptions": "",
    "DisplayName": "",
    "Address": "",
    "SecondaryAddress": "",
    "City": "",
    "State": "",
    "PostalCode": "",
    "Country": "",
    "Latitude": 40.59485,
    "Longitude": -73.96174,
    "LatLonQuality": 99,
    "BusinessLogoUrl": "",
    "BusinessUrl": "",
    "DisplayText": "",
    "PhoneNumber": "",
    "VenueGroup": 7,
    "VenueType": 0,
    "SubVenue": 0,
    "IndoorFlag": "",
    "OperatorDefined": "",
    "AccessPoints": [{
        "AccessPointId": "",
        "MACAddress": "",
        "DisplayFlag": "",
        "DisplayOptions": "",
        "Latitude": 40.59485,
        "Longitude": -73.96174,
        "Status": "Up",
        "OperatorDefined": "",
        "RoamingGroups": [{
            "GroupName": ""
        },
        {
            "GroupName": ""
        }],
        "Radios": [{
            "RadioId": "",
            "RadioFrequency": "",
            "RadioProtocols": [{
                "Protocol": ""
            }],
            "WifiConnections": [{
                "BSSID": "",
                "ServiceSets": [{
                    "SSID": "",
                    "SSID_Broadcasted": ""
                }]
            }]
        }]
    }]
}

我试图使用Json.NET JsonTextReader来实现这一点,但是我无法让读者在其缓冲区中包含整个位置,因为读者最初将拥有的流中的记录大小直到" RadioProtocols",这是通过对象的中途,当流到达对象的末尾时,读者已经放弃了对象的开始。

我尝试使用此功能的代码是

var ser = new JsonSerializer();
using (var reader = new JsonTextReader(new StreamReader(stream)))
{
    reader.SupportMultipleContent = true;

    while (reader.Read())
    {   
        if (reader.TokenType == JsonToken.StartObject && reader.Depth == 2)
        {                            
            do
            {
                reader.Read();                                
            } while (reader.TokenType != JsonToken.EndObject && reader.Depth == 2);

            var singleLocation = ser.Deserialize<Locations>(reader);
        }
    }
}

有关此的任何信息或替代方法将非常感谢。作为旁注,我们的客户发送信息的方式目前无法改变。

2 个答案:

答案 0 :(得分:0)

当读者位于要反序列化的对象的开头(在您的情况下为Locations数组中的条目)时,您只需调用ser.Deserialize<T>(reader)它就会起作用,前进到在该级别的对象的结束,没有进一步。因此,以下内容应迭代文件中的Location个对象,分别加载每个对象:

    public static IEnumerable<T> DeserializeNestedItems<T>(TextReader textReader)
    {
        var ser = new JsonSerializer();
        using (var reader = new JsonTextReader(textReader))
        {
            reader.SupportMultipleContent = true;

            while (reader.Read())
            {
                if (reader.TokenType == JsonToken.StartObject && reader.Depth == 2)
                {
                    var item = ser.Deserialize<T>(reader);
                    yield return item;
                }
            }
        }
    }

使用测试字符串的一个使用示例:

        Debug.Assert(DeserializeNestedItems<Location>(new StringReader(json)).Count() == 2); // No assert.

        var list = DeserializeNestedItems<Location>(new StringReader(json)).SelectMany(l => l.AccessPoints).Select(a => new { a.Latitude, a.Longitude }).ToList();

        Debug.WriteLine(JsonConvert.SerializeObject(list, Formatting.Indented));

哪个输出:

[
  {
    "Latitude": 40.59485,
    "Longitude": -73.96174
  },
  {
    "Latitude": 40.59485,
    "Longitude": -73.96174
  }
]

注意 - Location课程来自将您的JSON发布到http://json2csharp.com/

答案 1 :(得分:0)

感谢所有的帮助,我设法让它按照我想要的方式进行反序列化单个位置对象。

如果项目被转换为JObject,它将读入完整对象并对其进行反序列化,这可以循环以获得解决方案。

这是在

上解决的代码
while (reader.Read())
{
    if (reader.TokenType == JsonToken.StartObject && reader.Depth == 2)
    {
        location = JObject.Load(reader).ToObject<Location>();

        var lv = new LocationValidator(location, FootprintInfo.OperatorId, FootprintInfo.RoamingGroups, true);
        var vr = lv.IsValid();
        if (vr.Successful)
        {
            yield return location;
        }
        else
        {
            errors.Add(new Error(elNumber, location.LocationId, vr.Error.Field, vr.Error.Detail));
            if (errors.Count >= maxErrors)
            {
                yield break;
            }
        }

        ++elNumber;
    }
}