Bad request error when using MS Batch MT with segmented TUs

[Former user]

Original issue 415 created by @ysavourel on 2014-09-09T19:16:42.000Z:

It seems the MS MT connector batch translation step is not handling segmented TUs properly when sending the requests. We get 400 Bad Request in those case.

The work around is to move to the TU level, but we need to fix the normal case.
The SOAP message looks OK, but it may be related to the number of <string> as the counter goes for more strings than when the same TUs go by text container.

[Former user]

Comment 1. originally posted by @ysavourel on 2014-09-09T21:20:01.000Z:

It looks like the issue is a limitation on the the number of items in the array of the request: the API doc now says:
"Required. An array containing the texts for translation. All strings must be of the same language. The total of all texts to be translated must not exceed 10000 characters. The maximum number of array elements is 10."

so 10 strings max and no more than 10000 character total are the limits.

[Former user]

Comment 2. originally posted by @ysavourel on 2014-09-09T21:20:44.000Z:

The doc for reference: http://msdn.microsoft.com/en-us/library/ff512418.aspx

[Former user]

Comment 3. originally posted by @ysavourel on 2014-10-27T20:44:18.000Z:

Likely same issue reported here:
https://groups.yahoo.com/neo/groups/okapitools/conversations/messages/4445
https://groups.yahoo.com/neo/groups/okapitools/conversations/messages/4446

[Chase Tingley]

MS MT Connector - Overhaul of API implementation

Two functional changes:
- query(String) will now send the text it is passed directly, without
  HTML-escaping any inline codes.  This improves the behavior when
  dealing with content containing unparsed markup, which was
  previously garbled by the connector's call to toCodedHTML().
  batchQueryText() was added to mirror this behavior in batch mode.
  query(TextFragment) and batchQuery(List<TextFragment>) behavior as
  before.
- Improved batching functionality: the connector will now obey both
  the MS API's 10,000 character limit *and* its limit of 10 segments
  per batch, and will split up larger batch requests internally.
  This will reduce the burden on code that calls the connector,
  which previously had to do some of this batching itself. This
  fixes issue #415.

Additionally, this cleans up many aspects of the implementatation
and improves unittest coverage.

→ <<cset 3bee80115acd>>