omero-cli-transfer

An OMERO CLI plugin for creating and using transfer packets between OMERO servers.

Transfer packets contain objects and annotations. This project creates a zip file from an object (Project, Dataset, Image, Screen, Plate) containing all original files necessary to create the images in that object, plus an XML file detailing the links between entities, annotations and ROIs thereof.

The CLI plugin add the subcommand transfer, which in its turn has two further subcommands omero transfer pack and omero transfer unpack. Both subcommands (pack and unpack) will use an existing OMERO session created via CLI or prompt the user for parameters to create one.

Installation

tl;dr: if you have python>=3.8, a simple pip install omero-cli-transfer might do. We recommend conda, though.

omero-cli-transfer requires at least Python 3.8. This is due to ome-types requiring that as well; this package relies heavily on it, and it is not feasible without it.

Of course, this CAN be an issue, especially given omero-py officially only supports Python 3.6. However, it is possible to run omero-py in Python 3.8 or newer as well. Our recommended way to do so it using conda. With conda installed, you can do

conda create -n myenv -c conda-forge python=3.8 zeroc-ice=3.6.5
conda activate myenv
pip install omero-cli-transfer

It is possible to do the same thing without conda as long as your python/pip version is at least 3.8, but that will require locally building a wheel for zeroc-ice (which pip does automatically) - it is a process that can be anything from "completely seamless and without issues" to "I need to install every dependency ever imagined". Try at your own risk.

If you want optional RO-Crate exports, you can do

pip install omero-cli-transfer[rocrate]

instead.

Usage

omero transfer pack

Creates a transfer packet for moving objects between OMERO server instances.

The syntax for specifying objects is: <object>:<id> where <object> can be Image, Project, Dataset, Plate or Screen. Project is assumed if <object>: is omitted. A file path needs to be provided; a tar file with the contents of the packet will be created at the specified path.

Currently, only MapAnnotations, Tags, FileAnnotations and CommentAnnotations are packaged into the transfer pack. All kinds of ROI (except Masks) should work.

Note that, if you are packing a Plate or Screen, default OMERO settings prevent you from downloading Plates and you will generate an empty pack file if you do so. If you want to generate a pack file from these entities, you will need to set omero.policy.binary_access appropriately.

--zip packs the object into a compressed zip file rather than a tarball.

--barchive creates a package compliant with Bioimage Archive submission standards - see below for more detail.

--rocrate generates a RO-Crate compliant package with flat structure (all image files in a single folder). A JSON metadata file is added with basic information about the files (name, mimetype).

--simple creates a "human-readable" package; one folder per project or dataset is created and image files are placed according to where they came from in the OMERO server. Note that a package generated with this option is not guaranteed to work with unpack, though it often will.

--metadata allows you to specify which transfer metadata will be saved in transfer.xml as possible MapAnnotation values to the images. Defaults to image ID, timestamp, software version, source hostname, md5, source username, source group.

--plugin allows you to export omero data to a desired format by using an external plugin. See for example the arc plugin, which exports omero projects to ARC repositories.

--binaries allows to specify whether to archive binary data (e.g images, ROIs, FileAnnotations) or only create the transfer.xml. Default is all and will create the archive. With none, only the transfer.xml file is created, in which case the last cli argument is the path where the transfer.xml file will be written.

--ignore_errors gnores any download/export errors during the pack process, often the result of servers which do not allow Plate downloads (but will ignore any non-zero return code from omero download or omero export).

Examples:

omero transfer pack Image:123 transfer_pack.tar
omero transfer pack --zip Image:123 transfer_pack.zip
omero transfer pack Dataset:1111 /home/user/new_folder/new_pack.tar
omero transfer pack 999 tarfile.tar  # equivalent to Project:999
omero transfer pack --plugin arc Project:999 path/to/my/arc/repo
omero transfer pack --binaries none Dataset:1111 /home/user/new_folder/
omero transfer pack --binaries all Dataset:1111 /home/user/new_folder/new_pack.tar

omero transfer unpack

Unpacks an existing transfer packet, imports images/plates as orphans and uses the XML contained in the transfer packet to re-create links, annotations and ROIs.

Note that unpack needs to be able to identify the images it imports inequivocally; this can be a problem in case you have other images with the same clientPath (i.e. that were imported from the exact same location, including filename) and no annotations created by omero-cli-transfer. The most common case to generate this issue is an unpack that fails after the import step - the lingering images are not annotated correctly and a retry of the same unpack will use the same clientPath and cause issues. The best solution is cleaning up after failed unpacks.

--ln_s forces imports to use the transfer=ln_s option, in-place importing files. Same restrictions of regular in-place imports apply.

--output allows for specifying an optional output folder where the packet will be unzipped.

--folder allows the user to point to a previously-unpacked folder rather than a single file.

--merge will use existing Projects, Datasets and Screens if the current user already owns entities with the same name as ones defined in transfer.xml, effectively merging the "new" unpacked entities with existing ones.

--metadata allows you to specify which transfer metadata will be used from transfer.xml as MapAnnotation values to the images. Fields that do not exist on transfer.xml will be ignored. Defaults to image ID, timestamp, software version, source hostname, md5, source username, source group.

Examples:

omero transfer unpack transfer_pack.zip
omero transfer unpack --output /home/user/optional_folder --ln_s
omero transfer unpack --folder /home/user/unpacked_folder/

omero transfer prepare

Creates an XML from a folder with images.

Creates an XML file appropriate for usage with omero transfer unpack from a folder that contains image files, rather than a source OMERO server. This is intended as a first step on a bulk-import workflow, followed by using omero transfer unpack to complete an import.

Examples:

omero transfer prepare /home/user/folder_with_files

NOTE: please refer to optional requirement instructions below and consider that this feature is experimental!

Bioimage Archive submission contents

• Folder structure in the generated zip/tar follows project/dataset structure rather than original ManagedRepository folder structure, and instead of a transfer.xml file, a submission.tsv file is generated.

• submission.tsv file has:

  • one line per file being submitted, between Image files and FileAnnotation files;
  • a column indicating whether that file was originally an Image or FileAnnotation;
  • a "comment" column if any Image has a CommentAnnotation;
  • a column per key in a MapAnnotation inside the pack, with an empty value for all images but the ones with a MapAnnotation with that key; for those images, it has the value for that annotation;
  • a final original_omero_ids column listing all OMERO IDs associated to that file in the origin server: for images, that is all Image IDs that use that file, and for file annotations that is all Image IDs that had that FileAnnotation attached to them.

RO-Crate export format

• This requires an optional dependency on ro-crate-py that can be installed with pip install omero-cli-transfer[rocrate].
• Largely due to library limitations, current exports create a flat structure inside a zip file. For each image, name and mimetype are recorded. A ro-crate-metadata.json is added to the zip file.

omero transfer prepare optional requirements

• prepare requires bftools to work (in particular, we need to be able to run showinf). The easiest way to install this is by using conda; A simple conda install -c bioconda bftools on your conda environment should suffice.
• Note that this is a Java application. The conda package will install a JDK for you if necessary, but if you're installing it for yourself you'll need to make sure Java is available.
• This tool runs showinf in a subprocess and needs to be able to parse the output. This can be problematic if your stdout is not set to UTF-8; it can mess up special characters is e.g. measurement units and lead to XML validation errors. In addition to that, Java itself might output data in non-UTF-8 encodings, in which case it might be necessary to set JAVA_TOOL_OPTIONS=-Dfile.encoding=UTF8.

Plugin interface for external packers

External export formats can be used by omero-cli-transfer via plugin interface:

omero transfer pack --plugin my-exporter Dataset:111 path/to/my_export.zip

Example: ARC plugin

With the arc plugin, OMERO projects can be transferred to ARC repositories.

omero transfer pack --plugin arc Project:111 path/to/my/arc_repo

Plugins for omero-cli-transfer can be created by providing an entry point with group name omero_cli_transfer.pack.plugin. Entry points are defined in setup.py or pyproject.toml, , see e.g. the entrypoint definition of the arc plugin.

The entry point must be a function with following arguments:

• ome_object: The omero object wrapper to pack.
• destination_path: The export target path.
• tmp_path: The temporary path where downloaded images and transfer.xml are located.
• image_filenames_mapping: A dict that maps image ids to filenames.