Missing test changes.

See nexedi/jio@ff4846ab

Thanks to Thomas Gambier for the fix.

Ensure the thenable is correctly cancelled.

With this update, jiodev will provide the possibility to specify parameters of conversion from html to pdf using cloudooo and wkhtmltopdf.

/reviewed-on nexedi/jio!110

The array buffer length was higher than what is stored in indexedDB

Thanks to Alexandra Rogova for reporting the issue.

Use Blob.size instead...

If a document stop matching the replicate query, it will be considered as deleted.
But its attachments will still be reachable and previously prevented the remote document deletion.

Fix erp5 storage query handling

Fix query.

Improve IndexedDB storage.

Improve ajax in node.

- fixes `"\"a b\""`
  parsed to {simple query with value: '"a b"'}
  serialized to ` "\"a b\""`

/bug #20190603-E76ED6
https://nexedijs.erp5.net/#/bug_module/20190603-E76ED6

/reviewed-on !107

Thanks to Guillaume Royer <guillaume.royer@clearroad.io> for reporting
the issue.
See nexedi/jio!103

Try to reproduce the browser specifications.

sinon now correctly replace it only on demand.

Do not spy XMLHttpRequest, as it is node 100% compatible between node and browser.
Spy ajax calls instead.

Do not rely on storage not available in node (like localstorage).

Deprecate websql.

Add group_by parameter to allDocs.

Queries: Fix some parsing/serialization issues

- fixes query toString inconsistent output
- change serialization behavior
  ( a:b OR a:c ) -> a:( b OR c )
  a:( b OR c )   -> a:( b OR c )
- rename simpleQuerySetKey to querySetKey
  as it also applies on complex queries.

https://www.w3.org/TR/webdatabase/

Fixup linshare

Add LinshareStorage

LinshareStorage uses the linshare document name as jio key.

If multiple linshare documents have the same name, only the most recent modified is used.

Thanks to Vincent Bechu for his work on this storage.

Fix IndexedDB transaction handling.

Ease usage of ReplicateStorage.

Accumulate all operations done in a ReplicateReport object.

This ReplicateReport is returned by the repair method in any case (success or failure).

If an error occured while replicating a document, do not stop the full process and log the issue in the report.
Which means repair now always tries to replicate all documents even if one fails.

Each operation has a code to ease parsing it if needed.

By default, the report only contains error code (< 100).
It is possible to change the log level by using the 'report_level' storage parameter.

If the 'debug' parameter is set to true, codes will also be displayed in the console.log.

Open database is now wrapped into a promise which ensure that the database will be closed as soon as the promise is resolved.
This will allow to open multiple transactions if needed.

Transaction creation is now manage by a promise, which will abort the transaction if cancel is called.

The transaction will also be aborted if an error occurs during its callback function.

As transaction abort will reject all accesses, it is safe to use .then method inside the transaction callback.

Check the transaction error event to detect some browser internal errors (like QuotaExceededError).
Previously, such error was silently ignored, and the jIO client was not aware that the data were not correctly written.

All deletions are now correctly checked, and the method will wait for all deletion to be successfull before returning.

Improve putAttachment algorithm to make it update the previous value and delete only the useless blob chunks.

Reduce the usage of openCursor to not fetch the table content if not needed (like if the index key is enough for example).