`Resource`: fix a name when the resource is created, and refuse to change it later by BioCam · Pull Request #1228 · PyLabRobot/pylabrobot

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCam · 2026-08-31T14:32:15Z

A resource's name is its identifier: it is unique across the tree, it is what get_resource looks a resource up by, and it is the key serialize_all_state writes state under. Until now it could still be reassigned after the resource existed, and anything already named after it was left behind.

The problem

A well is named three times before it is usable - well_0_0 when it is constructed, well_A1 by create_ordered_items_2d, and finally plate_well_A1 by ItemizedResource.__init__, which prefixes every item it is handed with its own name. The last two are renames, and the last one writes the plate's name into all 96 of its wells.

Because the plate itself can still be renamed, those wells can be left describing a plate that no longer exists:

plate=Cor_96_wellplate_360ul_Fb(name="plate") # wells: plate_well_A1 ...plate.name="sample_plate"# wells: plate_well_A1 ...

Nothing raises, but get_resource("sample_plate_well_A1") finds nothing and the plate's state is serialized under keys that no longer correspond to anything in the tree.

This PR

the name setter now refuses, raising AttributeError rather than RuntimeError - that is what Python raises for a read-only property, so setattr and anything catching attribute errors behave as they would for any other immutable attribute.
create_equally_spaced_2d takes a name_for callable, so each item is constructed with the name it will keep rather than renamed afterwards.
create_ordered_items_2d takes name_prefix, the name of the resource that will hold the items, and passes it down. This matches create_homogeneous_resources, which already takes a name_prefix and already constructs each site with its final name. Carriers have named their children this way all along; this brings the plate side onto the same convention. The two differ only in separator - carriers join with -, items with _, which is what ItemizedResource used and what keeps every name unchanged.
ItemizedResource.__init__ no longer renames what it is given.
the 167 call sites pass name_prefix: 155 pass name_prefix=name, the name their holder is being created with, and 12 in tests pass a literal. Four sites that build items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly.
named() writes the field rather than going through the setter, since the copy it returns did not exist a moment ago and belongs to no tree; it is being built, not renamed.

Behaviour: names are unchanged, verified byte for byte across 342 resources on a loaded STARlet deck, and named() returns what it always did. The only change a caller can see is that assigning to name now raises.

…ange it later A name is what a resource is identified by - it is unique across a tree, it is what `get_resource` finds a resource by, and it is the key its state is serialized under. Anything named after it takes that name at construction, so a name that changed afterwards left those disagreeing with it. The setter now refuses, with `AttributeError` rather than `RuntimeError`, since that is what Python raises for a read-only property. `create_ordered_items_2d` built its items and then renamed them from `well_0_0` to `well_A1`. It now passes the name it wants to the constructor, via a `name_for` argument on `create_equally_spaced_2d`. No call site changes: the naming was internal to both. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

…aming them `ItemizedResource.__init__` renamed every item it was given, prefixing it with its own name, so a well became `plate_well_A1` only after the plate existed. That is the one rename a fixed name cannot allow. `create_ordered_items_2d` takes the holder's name as `name_prefix` and passes it to each item's constructor instead. The 169 call sites pass `name_prefix=name`, the name the holder is being created with. Four sites that build their items by hand - two Opentrons loaders, the STAR teaching rack, and one test - name them after their holder directly. `named()` writes the field rather than going through the setter: its copy did not exist a moment ago and belongs to no tree, so it is being built rather than renamed, and the setter cannot tell the two apart. Names are unchanged: 342 resources on a loaded STARlet deck, byte for byte. Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

rickwierenga · 2026-08-31T19:36:10Z

 def named(self, name: str) -> Self:
 """Return a copy of this resource with the given name."""
 new_resource = self.copy()
- new_resource.name = name
+ # The copy is new and in no tree, so this finishes building it rather than renaming anything.
+ new_resource._name = name


this is also good

…gration Brings in 17 commits, including PyLabRobot#1228, which moved item naming from `ItemizedResource.__init__` into the item helpers. `n_channel_pipettes` builds its mounting shafts through `create_ordered_items_2d` and did not pass `name_prefix`, so the shafts came out as `tipmountingshaft_A1` rather than `<pipette>_tipmountingshaft_A1`. Passing the pipette's name restores what they were called before.

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

BioCamand others added 2 commits August 31, 2026 15:10

BioCam force-pushed the resource-immutable-name branch from ec56fc9 to 1ccc0e7Compare August 31, 2026 14:39

BioCam requested a review from rickwierenga August 31, 2026 15:16

BioCam mentioned this pull request Aug 31, 2026
Scaling: Resource: remember which names a tree holds instead of re-reading it #1222
Merged

rickwierenga reviewed Aug 31, 2026
View reviewed changes

rickwierenga approved these changes Aug 31, 2026
View reviewed changes

BioCam merged commit 4b03d5e into mainAug 31, 2026
21 checks passed

BioCam deleted the resource-immutable-name branch August 31, 2026 21:42

Uh oh!

Conversation

BioCam commented Aug 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The problem

This PR

Uh oh!

rickwierengaAug 31, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

BioCam commented Aug 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The problem

This PR

Uh oh!

rickwierengaAug 31, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

BioCam commented Aug 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The problem

This PR

Uh oh!

rickwierengaAug 31, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

BioCam commented Aug 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

The problem

This PR

Uh oh!

rickwierengaAug 31, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

BioCam commented Aug 31, 2026• edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading

BioCam commented Aug 31, 2026•
edited
Loading