Reputation: 1357
I need to preform ETL like operations onto a JSON file that I receive inside of a .tar.gz. I have been able to successfully unzip and untar the file into a memory stream.
My issue is this, these files are so large that I see memory issues when trying to use various libraries to parse the file into objects. I have used suggestions from a few places to try not to read the whole file but they still involve throwing the whole file into memory it would seem.
https://www.newtonsoft.com/json/help/html/Performance.htm https://www.newtonsoft.com/json/help/html/SerializingJSONFragments.htm
What I would like to do, is asynchronously read a lite bit of the file, try to parse this read into objects, and then add to the appropriate queues for further processing. My hope is that I can strip these objects out of memory as quickly as possible.
Example JSON
{
"header" : {
"id" : 12345,
"datetime" : 1640423287060050040,
"version" : 1.0
},
"Reading" : [
{
"id" : 54321,
"units" : "fps",
"data" : [
{
"value" : 32,
"time" : 1630000000000000400
},
{
"value" : 32,
"time" : 1630000000000000400
},
{
"value" : 32,
"time" : 1630000000000000400
}
]
},
{
"id" : 765432,
"units" : "fps",
"data" : [
{
"value" : 21,
"time" : 1630000000000000400
},
{
"value" : 21,
"time" : 1630000000000000400
},
{
"value" : 21,
"time" : 1630000000000000400
}
]
}
]
}
This is memory inefficent due to the ReadToEndAsync()
using (var stream = _readFile(inFilePath)) //TODO Read async
{
if (stream == null || stream == StreamReader.Null)
{
throw new Exception("streamReader is null");
}
data = JObject.Parse(await stream.ReadToEndAsync());
}
}
Where im stuck
var memoryStream = UnTarGz.ExtractTarGzToStream(inFilePath);
memoryStream.Position = 0;
using (var streamReader = new StreamReader(memoryStream))
{
using (JsonReader reader = new JsonTextReader(streamReader))
{
var header = new Header();
while (await reader.ReadAsync(CancellationToken.None))
{
//if somehow I can detect header set parameter above as header data.
//else if its reading data build an object such that
// {
// header = header;
// readings = the reading data
// }
//
// add to queue
}
}
}
Upvotes: 0
Views: 871
Reputation: 4553
Jeroen provided an overview of the solution in the comments. At the time of writing my reputation is about 5% of his, so it took me a while to figure out what that actually meant :) but as I finally got it working I'll post the result here in case this is what you're looking for:
public static class Json
{
public static void Run()
{
using (Stream s = File.OpenRead(@"C:\Temp\input.json")) // This is your sample.
using (var sr = new StreamReader(s))
using (var reader = new JsonTextReader(sr))
{
var serializer = new JsonSerializer();
var readingCounter = 0;
// While there are tokens to be read, do it and update reader with token attributes.
while (reader.Read())
{
if (reader.TokenType == JsonToken.StartObject && reader.Path == "Header")
{
var header = serializer.Deserialize<Header>(reader);
Console.WriteLine(header);
}
if (reader.TokenType == JsonToken.StartObject && reader.Path == $"Reading[{readingCounter}]")
{
var currentReading = serializer.Deserialize<Reading>(reader);
readingCounter++;
// Process current reading object here (ex. "add to queue"):
Console.WriteLine(currentReading);
}
}
}
}
}
public class Header
{
public long Id { get; set; }
public long Datetime { get; set; }
public string Version { get; set; }
// For the example only:
public override string ToString()
{
return Id + ", " + Datetime + ", " + Version;
}
}
public class Reading
{
public int Id { get; set; }
public string Units { get; set; }
public List<ValueTime> Data { get; set; }
// For the example only:
public override string ToString()
{
return Id + ", " + Units + ", data count = " + Data.Count;
}
}
public class ValueTime
{
public int Value { get; set; }
public long Time { get; set; }
}
Upvotes: 1