Is there a way to overwrite duplicate cells when using extract? about gdspy

[heitzmann]

@peterxsu
It's not a simple matter of duplicate names that define identical subcells. Cells could have the same name and different contents. It could be dangerous for users to fall into this trap without knowing it.

In your case, if you know that cells with the same name are identical, the following code should give you an idea of what you have to do to merge the files:

import gdspy


def create_gdsii(n):
    fname = 'file{}.gds'.format(n)

    cell1 = gdspy.Cell('REFERENCED', True)
    cell1.add(gdspy.Rectangle((0, -0.5), (1, 0.5)))
    cell1.add(gdspy.Label('Cell from {} '.format(fname), (0, 0), 'e'))

    cell2 = gdspy.Cell('COPY{}'.format(n), True)
    cell2.add(gdspy.CellArray(cell1, 1, 1 + n, (0, 2)))

    lib = gdspy.GdsLibrary()
    lib.add([cell1, cell2])
    lib.write_gds(fname)
    return fname


# Create 2 example files
file1 = create_gdsii(1)
file2 = create_gdsii(2)

# Load first file
lib1 = gdspy.GdsLibrary(infile=file1)
# Extract all cells:
for cell in lib1.top_level():
    lib1.extract(cell)

# Load second file
lib2 = gdspy.GdsLibrary(infile=file2)

# Change references based on cell names
# WARNING: this is no guarantee that the cells are identical!
for cell in lib2.cell_dict.values():
    for element in cell.elements:
        if (isinstance(element, gdspy.CellArray) or
                isinstance(element, gdspy.CellReference)):
            name = element.ref_cell.name
            if name in gdspy.current_library.cell_dict:
                element.ref_cell = gdspy.current_library.cell_dict[name]

# Manually extract all cells skipping duplicate names
for name, cell in lib2.cell_dict.items():
    if name not in gdspy.current_library.cell_dict:
        gdspy.current_library.add(cell)

gdspy.LayoutViewer()

Note that, if you are using several files, I'd advise against using the global gdspy.current_library. Instead, you can work with a merged_lib as follows:

import gdspy


def create_gdsii(n):
    fname = 'file{}.gds'.format(n)

    cell1 = gdspy.Cell('REFERENCED', True)
    cell1.add(gdspy.Rectangle((0, -0.5), (1, 0.5)))
    cell1.add(gdspy.Label('Cell from {} '.format(fname), (0, 0), 'e'))

    cell2 = gdspy.Cell('COPY{}'.format(n), True)
    cell2.add(gdspy.CellArray(cell1, 1, 1 + n, (0, 2)))

    lib = gdspy.GdsLibrary()
    lib.add([cell1, cell2])
    lib.write_gds(fname)
    return fname


# Create 2 example files
file1 = create_gdsii(1)
file2 = create_gdsii(2)

# Load first file into merged library
merged_lib = gdspy.GdsLibrary(infile=file1)

# Load second file
lib2 = gdspy.GdsLibrary(infile=file2)

# Change references based on cell names
# WARNING: this is no guarantee that the cells are identical!
for cell in lib2.cell_dict.values():
    for element in cell.elements:
        if (isinstance(element, gdspy.CellArray) or
                isinstance(element, gdspy.CellReference)):
            name = element.ref_cell.name
            if name in merged_lib.cell_dict:
                element.ref_cell = merged_lib.cell_dict[name]

# Manually extract all cells skipping duplicate names
for name, cell in lib2.cell_dict.items():
    if name not in merged_lib.cell_dict:
        merged_lib.add(cell)

gdspy.LayoutViewer(merged_lib)

[peterxsu]

Thanks @heitzmann . Sorry for not seeing this earlier.

Is there a reason that add has the overwrite_duplicate functionality then, and that's not considered too dangerous to give as an option?

[heitzmann]

@peterxsu I see your point. The difference there is that when using GdsLibrary.add, the user is expected to be creating all cells in Gdspy, meaning that they know what they're doing. By giving this option in GdsLibrary.extract I fell the issues with the references could be more problematic, because they are not really explicit.

But I think you are right, if the somewhat dangerous option is already there, we might as well make it available everywhere. Implemented in 150945a