Join ORC parts and create CSV - hdfs

What is the simplest way to convert an existing ORC store in many parts on HDFS to one CSV file? I can use hive and/or spark, but I want to keep the memory requirements small, so I don't want to spark.read and coalesce.

I'm not sure but you can try to export from hive the ORC to an external table in CSV format, setting hive.merge.mapredfiles=true
(https://community.hortonworks.com/questions/92415/how-to-merge-reduce-task-output-to-produce-final-o.html)

Related

Why are tables segmented when exporting to parquet from AWS RDS

We use Python's boto3 library to execute start_export_task to trigger a RDS snapshot export (to S3). This successfully generates a directory in S3 that has a predicable, expected structure. Traversing down through that directory to any particular table directory (as in export_identifier/database_name/schema_name.table_name/) I see several .parquet files.
I download several of these files and convert them to pandas dataframes so I can look at them. They are all structured the same and seem to clearly be pieces of the same table. But they range in size from 100KB to 8MB in seemingly unpredictable size segments. Do these files/'pieces' of the table account for all its rows? Do they repeat/overlap at all? Why are they segmented so (seemingly) randomly? What parameters control this segmenting?
Ultimately I'm looking for documentation on this part of parquet folder/file structure. I've found plenty of information on how individual files are structured and partitioning. But I think this falls slightly outside of those topics.
You're not going to like this, but from AWS' perspective this is an implementation detail and according to the docs:
The file naming convention is subject to change. Therefore, when reading target tables we recommend that you read everything inside the base prefix for the table.
— docs
Most of the tools that work with Parquet don't really care about the number or file names of the parquet files. You just point something like Spark or Athena to the prefix of the table and it will read all the files and figure out how they fit together.
In the API there are also no parameters to influence this behavior. If you prefer a single file for aesthetic reasons or others, you could use something like a Glue Job to read the table prefixes, coalesce the data per table in a single file and write it to S3.

Can we set remove column names from s3 partition path and set path to values?

I am just curious, for Spark using Glue sinkFormat, is it possible to save the file as "2021/05/05/filename.parquet" and not as "year=2021/month=05/day=05/filename.parquet". I tried to play with 'writepath' but it works at record level and I believe it will break Spark's ability to save partitioned files.
This is not possible.
Partitioning drops the columns used for partitioning.
Spark uses directory structure for partition discovery and the correct structure so including column names is necessary for it to work.

Athena Query Results: Are they always strings?

I'm in the process of building new "ETL" pipelines with CTAS. Unfortunately, Quite often the CTAS query is too intensive which causes Athena to time out. As such, I use CTAS to create the initial table and populate with a small sample. I then write a script that queries the same table the CTAS was generated from (which is parquet format) for the remaining days that the CTAS couldn’t handle upfront. I write the output of these query results to the same directory that is holding the results of the CTAS query before repairing the table (to pick up new data). However, it seems to be a pretty clunky process for a number of reasons:
1) Query results written out with a standard SQL statements all end up being strings. For example, when I write out the number of DAUs (which is a count and cast to an int) the csv output is a string I.e. wrapped in “”.
Is it possible to write out Athena "query_results" (not the CTAS) as anything other than a string when in CSV format. The main problem with this is it means it can't be read back into the table produced by the CTAS since these column expect a bigint. This, of course, can be resolved with a lambda function but seems like a big overhead for something that should be trivial.
2) Can you put query results (not from CTAS) directly into parquet instead of CSV?
3) Is there any way to prevent metadata being generated with the query_results (not from CTAS). Again, it can be cleaned up with a lambda function, but it's just additional nonsense I need to handle.
Thanks in advance!
The data type of the result depends on the SQL used to create it and also on how you consume it. Based on your question I'm going to assume that you're creating a table using CTAS and that the output is CSV, and that you're then looking at the CSV data directly.
That CSV is going to have quotes in it, but that doesn't mean that it's not possible to read integer values as integers, and so on. Athena uses a schema-on-read approach, and as long as the serde can interpret a value as a particular type, that type will work as the type of the column.
If you query the table created by your CTAS operation you should get back integers for the integer columns.
Using CTAS you can also create output of different types, like JSON, Avro, Parquet, and ORC, that keep the type information. Just use the format property to select the output type.
I am a bit confused what you mean by your third question. With a normal query you get two files on S3, the data file and the metadata file, and they will be written to the output location given in the StartQueryExecution API call, but with a CTAS query you get the output data in a different location (given in the SQL) than the metadata file.
Are you actually using CTAS, or are you talking about the regular query result files?
Update after the question got clarified:
1) Athena is unfortunately unable to properly read it's own output in many situations. This is something that really surprises me that they never considered before launch. You might be able to set up a table that uses the regex serde.
2) No, unfortunately the only output of a regular query is CSV at this time.
3) No, the metadata is always written to the same prefix as the output.
I think your best bet is running multiple CTAS queries that select subsets of your source data, if there is a date column for example you could make one CTAS per month or some other time range that works. After the CTAS queries have completed you can move the result files into the same directory on S3 and create a final table that has that directory as its location.

How do I ensure that the AWS Glue crawler I've written is using the OpenCSV SerDe instead of the LazySimpleSerDe?

For context: I skimmed this previous question but was dissatisifed with the answer for two reasons:
I'm not writing anything in Python; in fact, I'm not writing any custom scripts for this at all as I'm relying on a crawler and not a Glue script.
The answer is not as complete as I require since it's just a link to some library.
I'm looking to leverage AWS Glue to accept some CSVs into a schema, and using Athena, convert that CSV table into multiple Parquet-formatted tables for ETL purposes. The data I'm working with has quotes embedded in it, which would be okay save for the fact that one record I have has a value of:
"blablabla","1","Freeman,Morgan","bla bla bla"
It seems that Glue is tripping over itself when it encounters the "Freeman,Morgan" piece of data.
If I use the standard Glue crawler, I get a table created with the LazySimpleSerDe, which truncates the record above in its column to:
"Freeman,
...which is obviously not desirable.
How do I force the crawler to output the file with the correct SerDe?
[Unpleasant] Constraints:
Looking to not accomplish this with a Glue script, since for that to work I believe I have to have a table beforehand, whereas the crawler will create the table on my behalf.
If I have to do this all through Amazon Athena, I'd feel like that would largely defeat the purpose but it's a tenable solution.
This is going to turn into a very dull answer, but apparently AWS provides its own set of rules for classifying if a file is a CSV.
To be classified as CSV, the table schema must have at least two
columns and two rows of data. The CSV classifier uses a number of
heuristics to determine whether a header is present in a given file.
If the classifier can't determine a header from the first row of data,
column headers are displayed as col1, col2, col3, and so on. The
built-in CSV classifier determines whether to infer a header by
evaluating the following characteristics of the file:
Every column in a potential header parses as a STRING data type.
Except for the last column, every column in a potential header has content that is fewer than 150 characters. To allow for a trailing
delimiter, the last column can be empty throughout the file.
Every column in a potential header must meet the AWS Glue regex requirements for a column name.
The header row must be sufficiently different from the data rows. To determine this, one or more of the rows must parse as other than
STRING type. If all columns are of type STRING, then the first row of
data is not sufficiently different from subsequent rows to be used as
the header.
I believed that I had met all of these requirements, given that the column names are wildly divergent from the actual data in the CSV, and ideally there shouldn't be much of an issue there.
However, in spite of my belief that it would satisfy the AWS Glue regex (which I can't find a definition for anywhere), I elected to move away from commas and to pipes instead. The data now loads as I expect it to.
Use glueContext.create_dynamic_frame_from_options() while converting csv to parquet and then run crawler over parquet data.
df = glueContext.create_dynamic_frame_from_options("s3", {"paths": [src]}, format="csv")
Default separator is ,
Default quoteChar is "
If you wish to change then check https://docs.aws.amazon.com/glue/latest/dg/aws-glue-programming-etl-format.html

What is the best approach to load data into Hive using NiFi?

I have started working with NiFi. I am working on a use case to load data into Hive. I get a CSV file and then I use SplitText to split the incoming flow-file into multiple flow-files(split record by record). Then I use ConvertToAvro to convert the split CSV file into an AVRO file. After that, I put the AVRO files into a directory in HDFS and I trigger the "LOAD DATA" command using ReplaceText + PutHiveQL processor.
I'm splitting the file record by record because to get the partition value(since LOAD DATA doesn't support dynamic partitioning). The flow looks like this:
GetFile (CSV) --- SplitText (split line count :1 and header line count : 1) --- ExtractText (Use RegEx to get partition fields' values and assign to attribute) --- ConvertToAvro (Specifying the Schema) --- PutHDFS (Writing to a HDFS location) --- ReplaceText (LOAD DATA cmd with partition info) --- PutHiveQL
The thing is, since I'm splitting the CSV file into each record at a time, it generates too many avro files. For ex, if the CSV file has 100 records, it creates 100 AVRO files. Since I want to get the partition values, I have to split them by one record at a time. I want to know is there any way, we can achieve this thing without splitting record by record. I mean like batching it. I'm quite new to this so I am unable to crack this yet. Help me with this.
PS: Do suggest me if there is any alternate approach to achieve this use case.
Are you looking to group the Avro records based on the partitions' values, one Avro file per unique value? Or do you only need the partitions' values for some number of LOAD DATA commands (and use a single Avro file with all the records)?
If the former, then you'd likely need a custom processor or ExecuteScript, since you'd need to parse, group/aggregate, and convert all in one step (i.e. for one CSV document). If the latter, then you can rearrange your flow into:
GetFile -> ConvertCSVToAvro -> PutHDFS -> ConvertAvroToJSON -> SplitJson -> EvaluateJsonPath -> ReplaceText -> PutHiveQL
This flow puts the entire CSV file (as a single Avro file) into HDFS, then afterwards it does the split (after converting to JSON since we don't have an EvaluateAvroPath processor), gets the partition value(s), and generates the Hive DDL statements (LOAD DATA).
If you've placed the file at the location where the hive table is reading the data using the puthdfs processor then you don't need to call the puthiveql processor. I am also new to this but I think you should leverage the schema-on-read capability of hive.