The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

NAME

MongoDB::Upgrading - Deprecations and behavior changes from v1 to v2

VERSION

version v2.2.2

DESCRIPTION

The v2 driver represents an evolutionary rather than revolutionary release, but with enough differences to justify a major version bump.

The most significant change in v2 is a switch away from the embedded BSON encoder/decoder to an external library, BSON and an optional optimization addon, BSON::XS. Many applications will continue to work unmodified, but some may need changes.

This document is intended to help developers update their code to take into account API changes from the v1 driver to the v2 driver.

RATIONALE

API Changes are never something to do lightly. Changes in the v2 driver were deemed necessary to achieve certain goals, all of which echo themes of the v1 driver release:

  • consistency – particularly with regards to Perl <-> BSON data conversion, the v2 driver provides a complete, consistently-designed set of BSON type wrappers, and significantly improved round-trip capabilities.

  • server compatibility – as the MongoDB server deprecates or removes functionality, the driver must be updated to match so that users don't develop apps around features that are going away.

  • portability – the switch to an external library that has both pure-Perl and XS optimized versions allows the MongoDB driver to support environments without a C compiler available.

INSTALLATION AND DEPENDENCY CHANGES

Installing v2 over v1

Because the v2 driver is pure-Perl capable (see below), its Perl installation directory is different. If upgrading, you need to be sure that the old version doesn't shadow the new one.

That's easy with cpanm:

    cpanm --uninst-shadows MongoDB

For the traditional CPAN client, you'll need to configure the make_install_arg config argument like this:

    $ perl -MCPAN -e shell
    cpan> o conf make_install_arg UNINST=1
    cpan> o conf commit
    cpan> install MongoDB

BSON library

The MongoDB driver uses a newer version of the BSON library. Previously, BSON was already required for BSON::Decimal128, so this is a version bump rather than an entirely new dependency.

Minimum Perl version

The MongoDB driver now requires Perl v5.10.1 or later. This provides better pure-Perl support, several core dependencies, and many fewer bugs involving Unicode and threads. While threads are discouraged, threads under Perl v5.8 were so broken that driver tests were regularly failing.

Pure-perl capable

The MongoDB driver can now be installed without needing a compiler. If a compiler is detected, additional XS-based dependencies will be added to the prerequisites list for improved performance. You can also specify PUREPERL_ONLY=1 as a Makefile.PL argument to disable compiler detection.

BSON BEHAVIOR CHANGES

For detailed information on handling MongoDB data types in Perl, see MongoDB::DataTypes. The following sections provide an overview of major changes from past versions.

MongoDB::BSON is removed

Code that customized behavior by instantiating this class will need to use BSON instead. Options are generally similar, though BSON provides much more flexibility.

New type wrapper classes

The BSON module provides a complete set of classes mapping to every BSON type. When decoding, these types will be returned for types that don't map by default to Perl types.

Code that uses ref to check documents returned from the database for legacy types (e.g. MongoDB::BSON::Regexp) will need to be updated for the new type wrappers.

Legacy type wrappers

All the legacy type wrappers have been updated to be subclasses of their corresponding BSON library equivalents. For example, MongoDB::BSON::Regexp is a subclass of BSON::Regex. Most of them are empty subclasses -- the BSON-library versions provide the same API -- but some have some additional constructor argument behaviors for backwards compatibility.

The BSON library knows how to encode legacy types, so code that uses legacy types for encoding values should be able to work without modification.

The legacy type wrappers will be removed in a future major version release of the driver.

Default date type decoding

The legacy driver defaulted to decoding the BSON date type as a DateTime object. Unfortunately, that type is very heavy-weight and slow to construct; it's a poor choice as a default as it inflicts that cost whether or not users ultimately need or want objects of that type.

The previously-deprecated dt_type configuration argument has been removed from MongoDB::MongoClient and the default date type of the BSON library is BSON::Time, which is extremely lightweight and provides convenience methods to convert to various popular time classes. It also works well with Time::HiRes for recording datetimes with millisecond precision.

Code that relied on date types being DateTime objects will need to convert via the as_datetime method of BSON::Time.

More consistent string/number heuristics

Depending on their history of use, non-reference Perl scalars may have both string and number representations internally and the MongoDB driver wasn't always clear on how it treated them. Moreover, this treatment could vary slightly by Perl version. The heuristics are now standardized as follows:

  • If the value has a valid double representation, it will be encoded to BSON as a double.

  • Otherwise, if the value has a valid integer interpretation, it will be encoded as either Int32 or Int64; the smallest type that the value fits will be used; a value that overflows will error.

  • Otherwise, the value will be encoded as a UTF-8 string.

The BSON library provides the prefer_numeric attribute to more aggressively coerce number-like strings that don't already have a numeric representation into a numeric form.

This is essentially the same as the legacy heuristic but some edge cases have been made consistent.

Type helper functions

To make it easy to use type wrappers (and to avoid unintentionally using a deprecated one), the BSON::Types module has a standard set of type helper functions:

    use BSON::Types ':all';

    $int32    = bson_int32(42);
    $time     = bson_time(); # now
    $ordered  = bson_doc( first => "John", last => "Doe );

NON-BSON BEHAVIOR CHANGES

run_command requires an ordered document

The MongoDB database uses the first key of the document provided to run_command as the name of the command. Due to Perl's hash order randomization, use of a hash reference with more than one key as an argument to run_command is not reliable. This restriction is now enforced. The argument must be a BSON::Doc object, a Tie::IxHash object, an array reference with an even number of keys, or a hash reference with a single key.

DEPRECATIONS

Count method on collections

The count method is deprecated.

The reasons for this change are as follows:

  • The performance and correctness characteristics of the count method could vary widely depending on whether or not a predicate is used.

  • The count method could be incorrect on sharded clusters during document migration between shards.

Many users are unaware of these considerations in the use of count. As any change to count could surprise users with unexpected differences in either performance or correctness, the count method has been replaced with two new API methods, which more directly convey performance and correctness expectations:

  • estimated_document_count takes no predicate; it does not work in transactions; performance is O(1).

  • count_documents takes a predicate (even if "empty", meaning count all documents); in can be used with or without transactions; performance is O(N) in the worst case.

NOTE: When upgrading from the deprecated count method, some legacy operators are not supported and must be replaced:

    +-------------+--------------------------------+
    | Legacy      | Modern Replacement             |
    +=============+================================+
    | $where      | $expr (Requires MongoDB 3.6+)  |
    +-------------+--------------------------------+
    | $near       | $geoWithin with $center        |
    +-------------+--------------------------------+
    | $nearSphere | $geoWithin with $centerSphere  |
    +-------------+--------------------------------+

Authentication

The MONGODB-CR authentication mechanism was deprecated in MongoDB server 3.6 and removed in MongoDB server 4.0. The Perl driver is deprecating MONGODB-CR, but will not remove it until it no longer supports older servers.

Query options

The following query options are deprecated:

  • maxScan -- deprecated in MongoDB server 4.0

  • modifiers -- the old $ prefixed modifiers have been replaced with explicit, equivalent options for find

  • snapshot -- deprecated in MongoDB server 4.0

MD5 checksum for GridFS files

The md5 field of GridFS documents is deprecated. Use of a checksum like MD5 has been redundant since MongoDB added write concern and MD5 itself is no longer considered a secure digest function. A future release will remove the use of MD5 entirely. In the meantime, users can disable MD5 digesting with the disable_md5 option in MongoDB::GridFSBucket.

Users who wish to continue storing a digest are encouraged to compute their own digest using a function of their choice and store it under a user-defined key in the metadata field of the file document.

Classes

These classes are superseded by type wrappers from BSON, as described earlier.

  • MongoDB::BSON::Binary

  • MongoDB::BSON::Regexp

  • MongoDB::Code

  • MongoDB::DBRef

  • MongoDB::OID

  • MongoDB::Timestamp

REMOVED FEATURES

Features deprecated in the v1 release have now been removed. Additionally, MongoDB::BSON has been removed in favor of BSON, as described earlier.

Configuration options

  • dt_type

  • query_timeout

  • sasl

  • sasl_mechanism

  • timeout

  • $MongoDB::BSON::char

  • $MongoDB::BSON::looks_like_number

Classes

  • MongoDB::BSON

  • MongoDB::GridFS

  • MongoDB::GridFS::File

Functions/Methods

  • From MongoDB - force_double, force_int

  • From MongoDB::BulkWrite and MongoDB::BulkWriteView - insert, update, remove, remove_one

  • From MongoDB::Collection - insert, batch_insert, remove, update, save, query, find_and_modify, get_collection, ensure_index, drop_indexes, drop_index, get_index, validate

  • From MongoDB::Database - eval, last_error, get_gridfs

  • From MongoDB::CommandResult - result

  • From MongoDB::Cursor - slave_okay, count

AUTHORS

  • David Golden <david@mongodb.com>

  • Rassi <rassi@mongodb.com>

  • Mike Friedman <friedo@friedo.com>

  • Kristina Chodorow <k.chodorow@gmail.com>

  • Florian Ragwitz <rafl@debian.org>

COPYRIGHT AND LICENSE

This software is Copyright (c) 2020 by MongoDB, Inc.

This is free software, licensed under:

  The Apache License, Version 2.0, January 2004