Adam Hooper created ARROW-7281:
----------------------------------

             Summary: AdaptiveIntBuilder::length() does not consider 
pending_pos_.
                 Key: ARROW-7281
                 URL: https://issues.apache.org/jira/browse/ARROW-7281
             Project: Apache Arrow
          Issue Type: Bug
          Components: C++
    Affects Versions: 0.15.1
            Reporter: Adam Hooper


{code:c++}
arrow::AdaptiveIntBuilder builder(arrow::default_memory_pool());
builder.Append(1);
std::cout << builder.length() << std::endl;
{code}

Expected output: {{1}}
Actual output: {{0}}

I imagine this regression came with https://github.com/apache/arrow/pull/3040

My use case: I'm building a JSON parser that appends "records" (JSON Objects 
mapping key=>value) to Arrow columns (each key gets an ArrayBuilder). Not all 
JSON Objects contain all keys; so {{builder.Append()}} isn't always called. So 
on a subsequent row, I want to add nulls for every append that was skipped: 
{{builder.AppendNulls(builder.length() - row); builder.Append(value)}}. This 
fails because {{builder.length()}} is wrong.

Annoying but simple workaround: I maintain a separate {{length}} value 
alongside {{builder}}.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to