I'm trying to write output from a scalding flow in json form, and reading it in Spark. This is working fine, except if the json contains strings with new lines. The output is one json object per line, and newlines in a value on the json is causing one bit of json to be fragmented across two lines. As such, when I read lines into Spark, I can't deserialize some of them. Is there a standard way to deal with this?
Related
Background: I want to store a dict object in json format that has say, 2 entries:
(1) Some object that describes the data in (2). This is small data mostly definitions, parameters that control, etc. and things (maybe called metadata) that one would like to read before using the actual data in (2). In short, I want good human readability of this portion of the file.
(2) The data itself is a large chunk- should more like machine readable (no need for human to gaze over it on opening the file).
Problem: How to specify some custom indent, say 4 to the (1) and None to the (2). If I use something like json.dump(data, trig_file, indent=4) where data = {'meta_data': small_description, 'actual_data': big_chunk}, meaning the large data will have a lot of whitespace making the file large.
Assuming you can append json to a file:
Write {"meta_data":\n to the file.
Append the json for small_description formatted appropriately to the file.
Append ,\n"actual_data":\n to the file.
Append the json for big_chunk formatted appropriately to the file.
Append \n} to the file.
The idea is to do the json formatting out the "container" object by hand, and using your json formatter as appropriate to each of the contained objects.
Consider a different file format, interleaving keys and values as distinct documents concatenated together within a single file:
{"next_item": "meta_data"}
{
"description": "human-readable content goes here",
"split over": "several lines"
}
{"next_item": "actual_data"}
["big","machine-readable","unformatted","content","here","....."]
That way you can pass any indent parameters you want to each write, and you aren't doing any serialization by hand.
See How do I use the 'json' module to read in one JSON object at a time? for how one would read a file in this format. One of its answers wisely suggests the ijson library, which accepts a multiple_values=True argument.
Issue
I recently encountered a challenge in Azure Data Lake Analytics when I attempted to read in a Large UTF-8 JSON Array file and switched to HDInsight PySpark (v2.x, not 3) to process the file. The file is ~110G and has ~150m JSON Objects.
HDInsight PySpark does not appear to support Array of JSON file format for input, so I'm stuck. Also, I have "many" such files with different schemas in each containing hundred of columns each, so creating the schemas for those is not an option at this point.
Question
How do I use out-of-the-box functionality in PySpark 2 on HDInsight to enable these files to be read as JSON?
Thanks,
J
Things I tried
I used the approach at the bottom of this page:
from Databricks that supplied the below code snippet:
import json
df = sc.wholeTextFiles('/tmp/*.json').flatMap(lambda x: json.loads(x[1])).toDF()
display(df)
I tried the above, not understanding how "wholeTextFiles" works, and of course ran into OutOfMemory errors that killed my executors quickly.
I attempted loading to an RDD and other open methods, but PySpark appears to support only the JSONLines JSON file format, and I have the Array of JSON Objects due to ADLA's requirement for that file format.
I tried reading in as a text file, stripping Array characters, splitting on the JSON object boundaries and converting to JSON like the above, but that kept giving errors about being unable to convert unicode and/or str (ings).
I found a way through the above, and converted to a dataframe containing one column with Rows of strings that were the JSON Objects. However, I did not find a way to output only the JSON Strings from the data frame rows to an output file by themselves. The always came out as
{'dfColumnName':'{...json_string_as_value}'}
I also tried a map function that accepted the above rows, parsed as JSON, extracted the values (JSON I wanted), then parsed the values as JSON. This appeared to work, but when I would try to save, the RDD was type PipelineRDD and had no saveAsTextFile() method. I then tried the toJSON method, but kept getting errors about "found no valid JSON Object", which I did not understand admittedly, and of course other conversion errors.
I finally found a way forward. I learned that I could read json directly from an RDD, including a PipelineRDD. I found a way to remove the unicode byte order header, wrapping array square brackets, split the JSON Objects based on a fortunate delimiter, and have a distributed dataset for more efficient processing. The output dataframe now had columns named after the JSON elements, inferred the schema, and dynamically adapts for other file formats.
Here is the code - hope it helps!:
#...Spark considers arrays of Json objects to be an invalid format
# and unicode files are prefixed with a byteorder marker
#
thanksMoiraRDD = sc.textFile( '/a/valid/file/path', partitions ).map(
lambda x: x.encode('utf-8','ignore').strip(u",\r\n[]\ufeff")
)
df = sqlContext.read.json(thanksMoiraRDD)
I'm working on some Python code for my local billiard hall and I'm running into problems with JSON encoding. When I dump my data into a file I obviously get all the data in a single line. However, I want my data to be dumped into the file following the format that I want. For example (Had to do picture to get point across),
My custom JSON format
. I've looked up questions on custom JSONEncoders but it seems they all have to do with datatypes that aren't JSON serializable. I never found a solution for my specific need which is having everything laid out in the manner that I want. Basically, I want all of the list elements to on a separate row but all of the dict items to be in the same row. Do I need to write my own custom encoder or is there some other approach I need to take? Thanks!
I have a sequence file containing multiple json records. I want to send every json record to a function . How can I extract one json record at a time?
Unfortunately there is no standard way to do this.
Unlike YAML which has a well-defined way to allow one file contain multiple YAML "documents", JSON does not have such standards.
One way to solve your problem is to invent your own "object separator". For example, you can use newline characters to separate adjacent JSON objects. You can tell your JSON encoder not to output any newline characters (by forcing escaping it into \ and n). As long as your JSON decoder is sure that it will not see any newline character unless it separates two JSON objects, it can read the stream one line at a time and decode each line.
It has also been suggested that you can use JSON arrays to store multiple JSON objects, but it will no longer be a "stream".
You can read content of your sequence files to RDD[String] and convert it to Spark Dataframe.
val seqFileContent = sc
.sequenceFile[LongWritable, BytesWritable](inputFilename)
.map(x => new String(x._2.getBytes))
val dataframeFromJson = sqlContext.read.json(seqFileContent)
When using crossfilter (for example for dc.js), do I always need to transform my data to a flat JSON for input?
Flat JSON data when reading from AJAX requests tend to be a lot larger than it needs to be (in comparison to for example nested JSON, value to array or CSV data).
Is there an API available which can read in other types than flat json? Are there plans to add those?
I would like to avoid to let the client transform the data before using it.