[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [DISCUSS] Unification of Hadoop related IO modules

+1 for going with option 3.

On Wed, Sep 12, 2018 at 8:51 AM Robert Bradshaw <robertwb@xxxxxxxxxx> wrote:
On Wed, Sep 12, 2018 at 5:27 PM Alexey Romanenko <aromanenko.dev@xxxxxxxxx> wrote:
Thank you everybody for your feedback!

I think we can conclude that the most popular option, according to discussion above, is number 3. Not sure if we need to do a separate vote for that but, please, let me know if we need.

So, for now, I’d split a work into the following steps:
a) Create new module "hadoop-mapreduce-format” which implements support for MapReduce OutputFormat through new HadoopMapreduceFormat.Write class. For that, I just need to change a bit my already created PR 6306 that I added recently (renaming of module and class names).
b) Move all source and test classes of “hadoop-input-format” into the module "hadoop-mapreduce-format” and create new class HadoopMapreduceFormat.Read there to support MapReduce InputFormat.
c) Make old HadoopInputFormat.Read (in old “hadoop-input-format” module) deprecated and as proxy class to newly created HadoopMapreduceFormat.Read (to keep API compatibility)

Sounds like a great plan. 
These 3 steps should be performed and completed within one release cycle (approx. in 2.8). For steps “b” and “c” I’d create another PR to avoid having a huge commit if it will include step “a” as well.

Big +1. 
Then, in next release after:
d) Remove completely module “hadoop-input-format”  (approx. in 2.9). 

I don't think we'd be able to remove this until 3.0. 

I think we technically we can remove HadoopInputFormat before 3.0 since it's marked as experimental [1] but I'd suggest keeping it deprecated for at least two releases (3 months) before removal. Not sure if we have a policy on this.


Other two Hadoop modules (common and file-system) we leave as it is.

I hope that this a correct summary of what community decided and I can move forward. 

Sounds good. 
Please, let me know if there any objections against this plan or other suggestions.

On 11 Sep 2018, at 16:08, Thomas Weise <thw@xxxxxxxxxx> wrote:

I'm in favor of a combination of 2) and 3): New module "hadoop-mapreduce-format" ("hadoop-format" does not sufficiently qualify what it is). Turn existing " hadoop-input-format" into a proxy for new module for backward compatibility (marked deprecated and removed in next major version).

I don't think everything "Hadoop" should be merged, purpose and usage is just too different. As an example, the Hadoop file system abstraction itself has implementation for multiple other systems and is not limited to HDFS.

On Tue, Sep 11, 2018 at 8:47 AM Alexey Romanenko <aromanenko.dev@xxxxxxxxx> wrote:
For now, you can’t write with Hadoop MapReduce OutputFormat. However, you can use FileIO or TextIO to write to HDFS, these IOs support different file systems.

On 11 Sep 2018, at 11:11, dharmendra pratap singh <dharmendra0393@xxxxxxxxx> wrote:

Hello Team,
Does this mean, as of today we can read from Hadoop FS but can't write to Hadoop FS using Beam HDFS API ?


On Thu, Sep 6, 2018 at 8:54 PM Alexey Romanenko <aromanenko.dev@xxxxxxxxx> wrote:
Hello everyone,

I’d like to discuss the following topic (see below) with community since the optimal solution is not clear for me.

There is Java IO module, called “hadoop-input-format”, which allows to use MapReduce InputFormat implementations to read data from different sources (for example, org.apache.hadoop.mapreduce.lib.db.DBInputFormat). According to its name, it has only “Read" and it's missing “Write” part, so, I'm working on “hadoop-output-format” to support MapReduce OutputFormat (PR 6306). For this I created another module with this name. So, in the end, we will have two different modules “hadoop-input-format” and “hadoop-output-format” and it looks quite strange for me since, afaik, every existed Java IO, that we have, incapsulates Read and Write parts into one module. Additionally, we have “hadoop-common” and “hadoop-file-system” as other hadoop-related modules. 

Now I’m thinking how it will be better to organise all these Hadoop modules better. There are several options in my mind: 

1) Add new module “hadoop-output-format” and leave all Hadoop modules “as it is”. 
Pros: no breaking changes, no additional work 
Cons: not logical for users to have the same IO in two different modules and with different names.

2) Merge “hadoop-input-format” and “hadoop-output-format” into one module called, say, “hadoop-format” or “hadoop-mapreduce-format”, keep the other Hadoop modules “as it is”.
Pros: to have InputFormat/OutputFormat in one IO module which is logical for users
Cons: breaking changes for user code because of module/IO renaming 

3) Add new module “hadoop-format” (or “hadoop-mapreduce-format”) which will include new “write” functionality and be a proxy for old “hadoop-input-format”. In its turn, “hadoop-input-format” should become deprecated and be finally moved to common “hadoop-format” module in future releases. Keep the other Hadoop modules “as it is”.
Pros: finally it will be only one module for hadoop MR format; changes are less painful for user
Cons: hidden difficulties of implementation this strategy; a bit confusing for user 

4) Add new module “hadoop” and move all already existed modules there as submodules (like we have for “io/google-cloud-platform”), merge “hadoop-input-format” and “hadoop-output-format” into one module. 
Pros: unification of all hadoop-related modules
Cons: breaking changes for user code, additional complexity with deps and testing

5) Your suggestion?..

My personal preferences are lying between 2 and 3 (if 3 is possible). 

I’m wondering if there were similar situations in Beam before and how it was finally resolved. If yes then probably we need to do here in similar way.
Any suggestions/advices/comments would be very appreciated.