Initial implementation of reprocessing xia2.multiplex jobs via SynchWeb by amyjaynethompson · Pull Request #384 · DiamondLightSource/python-dlstbx

amyjaynethompson · 2026-07-06T14:54:37Z

This PR provides an initial implementation to reprocess xia2.multiplex jobs. In a first instance, it might make sense to limit this feature to reprocessing existing jobs with different command line parameters. This is to keep the bookkeeping in SynchWeb as clear as possible.

This PR introduces a separate trigger function, because much of the main multiplex trigger function is unnecessary for a reprocessing job under this specification.

Examples of features of main trigger functions that don't work in this context:

loop through all sample groups associated with a DCID (for reprocessing, really only want to do a single sample group)
trigger every dataset vs at end of sample group (not really a consideration for reprocessing)
wait for datasets to complete (we know they've completed because it is being reprocessed)
check for space group compatibility, min number datasets, etc (we know this is fine because it has a parent job)
find files from xia2-dials jobs (we can get this directly from the parent job)

Instead, an ISPyB query can get the information required based on the parent job.

In SynchWeb, the new job is assigned a new processingJobId in ISPyB, and user defined parameters are given processingJobParameterIds. One of the processingJobParameters is the autoProcScalingId of the original multiplex job. From here, it is straightforward to query the database to find the DCIDs contained in the original job (as well as which sample group, and the data paths included).

This has been tested using ispyb-dev-1 which now has a multiplex reprocessing button. Using this interface submits the parameters to the ISPyB database. A new multiplex recipe is then called with zocalo.go -n -r trigger-multiplex -s ispyb_process=<processingJobId> where the processingJobId is the new job registered by SynchWeb (and linked to the user defined parameters).

…m parent job

pblowey · 2026-08-06T14:38:05Z

+    cchalf_filtering_method: Optional[str] = None
+    sd_cutoff: Optional[float] = None
+    image_group_size: Optional[int] = None
+    scaling_id: Optional[int] = None


scaling_id should be a required parameter as reprocessing will not work without it

Very true - have removed Optional

pblowey · 2026-08-07T16:40:17Z

+        # First, use the input scaling ID to find the DCIDs of all multiplex jobs as well as the sample (group) id value
+
+        query = (
+            session.query(ProcessingJobImageSweep, ProcessingJobParameter)
+            .join(
+                AutoProcProgram,
+                AutoProcProgram.processingJobId
+                == ProcessingJobImageSweep.processingJobId,
+            )
+            .join(
+                ProcessingJobParameter,
+                ProcessingJobParameter.processingJobId
+                == AutoProcProgram.processingJobId,
+            )
+            .join(
+                AutoProc,
+                AutoProc.autoProcProgramId == AutoProcProgram.autoProcProgramId,
+            )
+            .join(AutoProcScaling, AutoProcScaling.autoProcId == AutoProc.autoProcId)
+            .where(AutoProcScaling.autoProcScalingId == parameters.scaling_id)
+            .options(
+                Load(ProcessingJobImageSweep).load_only(
+                    ProcessingJobImageSweep.dataCollectionId,
+                    raiseload=True,
+                ),
+                Load(ProcessingJobParameter).load_only(
+                    ProcessingJobParameter.parameterKey,
+                    ProcessingJobParameter.parameterValue,
+                    raiseload=True,
+                ),
            )
+        )

-        return {"success": True, "return_value": jobids}
+        found_dcids = set()
+        group = None
+        found_data_files = set()
+
+        for dc, dc_params in query.all():
+            found_dcids.add(dc.dataCollectionId)
+            if "sample_id" == dc_params.parameterKey:
+                group = ("sample_id", dc_params.parameterValue)
+            elif "sample_group_id" == dc_params.parameterKey:
+                group = ("sample_group_id", dc_params.parameterValue)
+            elif "data" == dc_params.parameterKey:
+                found_data_files.add(dc_params.parameterValue)
+
+        # Set parameters
+
+        dcids = list(found_dcids)
+        data_files = list(found_data_files)
+
+        self.log.info(f"Found {dcids} corresponding to {group}")
+
+        self.log.debug(f"Found data files {data_files}")
+
+        job_parameters: list[tuple[str, str]] = [
+            ("data", files) for files in data_files
+        ]
+        if group:
+            job_parameters.append(group)
+


This could be simplified a lot by splitting into two queries. The issue with the single query is that you effectively get a duplicate of the processingJobParameters for each dcid. If you split it into two queries, the handling of the parameters becomes much easier.

Suggested change

# First, use the input scaling ID to find the DCIDs of all multiplex jobs as well as the sample (group) id value

query = (

session.query(ProcessingJobImageSweep, ProcessingJobParameter)

.join(

AutoProcProgram,

AutoProcProgram.processingJobId

== ProcessingJobImageSweep.processingJobId,

)

.join(

ProcessingJobParameter,

ProcessingJobParameter.processingJobId

== AutoProcProgram.processingJobId,

)

.join(

AutoProc,

AutoProc.autoProcProgramId == AutoProcProgram.autoProcProgramId,

)

.join(AutoProcScaling, AutoProcScaling.autoProcId == AutoProc.autoProcId)

.where(AutoProcScaling.autoProcScalingId == parameters.scaling_id)

.options(

Load(ProcessingJobImageSweep).load_only(

ProcessingJobImageSweep.dataCollectionId,

raiseload=True,

),

Load(ProcessingJobParameter).load_only(

ProcessingJobParameter.parameterKey,

ProcessingJobParameter.parameterValue,

raiseload=True,

),

)

)

return {"success": True, "return_value": jobids}

found_dcids = set()

group = None

found_data_files = set()

for dc, dc_params in query.all():

found_dcids.add(dc.dataCollectionId)

if "sample_id" == dc_params.parameterKey:

group = ("sample_id", dc_params.parameterValue)

elif "sample_group_id" == dc_params.parameterKey:

group = ("sample_group_id", dc_params.parameterValue)

elif "data" == dc_params.parameterKey:

found_data_files.add(dc_params.parameterValue)

# Set parameters

dcids = list(found_dcids)

data_files = list(found_data_files)

self.log.info(f"Found {dcids} corresponding to {group}")

self.log.debug(f"Found data files {data_files}")

job_parameters: list[tuple[str, str]] = [

("data", files) for files in data_files

]

if group:

job_parameters.append(group)

# Get the job parameters from the existing multiplex job.

job_parameters: list[tuple[str, str]] = (

session.query(ProcessingJobParameter.parameterKey, ProcessingJobParameter.parameterValue)

.select_from(AutoProcProgram)

.join(

ProcessingJobParameter,

ProcessingJobParameter.processingJobId

== AutoProcProgram.processingJobId,

)

.join(

AutoProc,

AutoProc.autoProcProgramId == AutoProcProgram.autoProcProgramId,

)

.join(AutoProcScaling, AutoProcScaling.autoProcId == AutoProc.autoProcId)

.where(AutoProcScaling.autoProcScalingId == parameters.scaling_id)

).all()

# Get the dataCollectionIds used in the existing multiplex job

parent_job_dcids: list[tuple[int]] = (

session.query(ProcessingJobImageSweep.dataCollectionId)

.join(

AutoProcProgram,

AutoProcProgram.processingJobId

== ProcessingJobImageSweep.processingJobId,

)

.join(

ProcessingJobParameter,

ProcessingJobParameter.processingJobId

== AutoProcProgram.processingJobId,

)

.join(

AutoProc,

AutoProc.autoProcProgramId == AutoProcProgram.autoProcProgramId,

)

.join(AutoProcScaling, AutoProcScaling.autoProcId == AutoProc.autoProcId)

.where(AutoProcScaling.autoProcScalingId == parameters.scaling_id)

).all()

dcids: list[int] = [row[0] for row in parent_job_dcids]

GitHub was grumpy about just applying this suggestion, but have now split the queries into two :)

pblowey · 2026-08-07T16:42:34Z

+        if parameters.spacegroup:
+            job_parameters.append(("spacegroup", parameters.spacegroup))
+
+        if parameters.d_min:
+            job_parameters.append(("d_min", str(parameters.d_min)))
+
+        if (
+            parameters.diffraction_plan_info
+            and parameters.diffraction_plan_info.anomalousScatterer
+        ):
+            job_parameters.extend(
+                [
+                    ("anomalous", "true"),
+                    ("absorption_level", "high"),
+                ]
+            )
+        if output_clusters:
+            job_parameters.extend(
+                [
+                    ("clustering.method", "coordinate"),
+                    ("clustering.output_clusters", "true"),
+                ]
+            )
+
+        # Unlike regular multiplex, filtering allowed regardless of beamline
+
+        if parameters.apply_cchalf_filtering:
+            job_parameters.append(("filtering.method", "deltacchalf"))
+            if parameters.cchalf_filtering_method:
+                job_parameters.append(
+                    ("deltacchalf.mode", parameters.cchalf_filtering_method)
+                )
+            if parameters.image_group_size:
+                job_parameters.append(
+                    ("deltacchalf.group_size", str(parameters.image_group_size))
+                )
+            if parameters.sd_cutoff:
+                job_parameters.append(
+                    ("deltacchalf.stdcutoff", str(parameters.sd_cutoff))
+                )


Some of this code, to append new parameters to the job parameters feels quite repetitive. It's not a huge issue but, particularly for the more simple additions, you could try doing this inside a loop to avoid having multiple if statements.

Have put this into a loop - for simplicity in the loop I've kept the parameter name and then edited the wrappers to use the proper Phil parameter (using same idea as space group and d_min)

amyjaynethompson · 2026-08-12T09:23:47Z

Tidied up code - local tests seem to work - to be merged and tested more during next shutdown once SynchWeb changes also merged.

…arent program

amyjaynethompson added 4 commits June 22, 2026 16:01

inital test of separate multiplex reprocessing trigger function

25e981d

ignore anomalous flags in multiplex filtering as will be imported fro…

317823d

…m parent job

separate multiplex reprocessing trigger to escape unnecessary logic

45424b4

fix jobid return

3ec92e8

amyjaynethompson requested a review from pblowey July 6, 2026 14:54

pblowey reviewed Aug 7, 2026

View reviewed changes

split query, and tidy parameter registration

9d12912

amyjaynethompson marked this pull request as ready for review August 12, 2026 09:22

amyjaynethompson added 2 commits August 12, 2026 10:28

remove beamline / clusters from parameters as currently taking from p…

3b5e4cd

…arent program

updating doc string from previous commit

d96c96e

Navigation Menu

Uh oh!

Initial implementation of reprocessing xia2.multiplex jobs via SynchWeb - #384

Initial implementation of reprocessing xia2.multiplex jobs via SynchWeb#384
amyjaynethompson wants to merge 7 commits into
mainfrom
multiplex_reprocessing

amyjaynethompson commented Jul 6, 2026

Uh oh!

pblowey Aug 6, 2026

Uh oh!

amyjaynethompson Aug 12, 2026

Uh oh!

pblowey Aug 7, 2026

Uh oh!

amyjaynethompson Aug 12, 2026

Uh oh!

pblowey Aug 7, 2026

Uh oh!

amyjaynethompson Aug 12, 2026

Uh oh!

amyjaynethompson commented Aug 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

amyjaynethompson commented Jul 6, 2026

Uh oh!

pblowey Aug 6, 2026

Choose a reason for hiding this comment

Uh oh!

amyjaynethompson Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

pblowey Aug 7, 2026

Choose a reason for hiding this comment

Uh oh!

amyjaynethompson Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

pblowey Aug 7, 2026

Choose a reason for hiding this comment

Uh oh!

amyjaynethompson Aug 12, 2026

Choose a reason for hiding this comment

Uh oh!

amyjaynethompson commented Aug 12, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants