|
| 1 | +# How do I ... |
| 2 | + |
| 3 | +## ... create datacubes from a directory of GeoTIFFs |
| 4 | + |
| 5 | +Files with GeoTIFF format cannot be opened directly by `zappend` unless |
| 6 | +you add [rioxarray](https://corteva.github.io/rioxarray/) to your |
| 7 | +Python environment. |
| 8 | + |
| 9 | +Then write your own [slice source](guide.md#slice-sources) and |
| 10 | +use configuration setting [`slice_source`](config.md#slice_source): |
| 11 | + |
| 12 | +```python |
| 13 | +import glob |
| 14 | +import numpy as np |
| 15 | +import rioxarray as rxr |
| 16 | +import xarray as xr |
| 17 | +from zappend.api import zappend |
| 18 | + |
| 19 | +def get_dataset_from_geotiff(tiff_path): |
| 20 | + ds = rxr.open_rasterio(tiff_path) |
| 21 | + # Add missing time dimension |
| 22 | + slice_time = get_slice_time(tiff_path) |
| 23 | + slice_ds = ds.expand_dims("time", axis=0) |
| 24 | + slice_ds.coords["time"] = xr.Dataset(np.array([slice_time]), dims="time") |
| 25 | + try: |
| 26 | + yield slice_ds |
| 27 | + finally: |
| 28 | + ds.close() |
| 29 | + |
| 30 | +zappend(sorted(glob.glob("inputs/*.tif")), |
| 31 | + slice_source=get_dataset_from_geotiff, |
| 32 | + target_dir="output/tif-cube.zarr") |
| 33 | +``` |
| 34 | + |
| 35 | +In the example above, function `get_slice_time()` returns the time label |
| 36 | +of a given GeoTIFF file as a value of type `np.datetime64`. |
| 37 | + |
| 38 | +## ... create datacubes from datasets without append dimension |
| 39 | + |
| 40 | +`zappend` expects the append dimension to exist in slice datasets and |
| 41 | +expects that at least one variable exists that makes use of that dimension. |
| 42 | +For example, if you are appending spatial 2-d images with dimensions x and y |
| 43 | +along a dimension time, you need to first expand the images into the time |
| 44 | +dimension. Here the 2-d image dataset is called `image_ds` and `slice_time` |
| 45 | +is its associated time value of type `np.datetime64`. |
| 46 | + |
| 47 | +```python |
| 48 | +slice_ds = image_ds.expand_dims("time", axis=0) |
| 49 | +slice_ds.coords["time"] = xr.Dataset(np.array([slice_time]), dims="time") |
| 50 | +``` |
| 51 | + |
| 52 | +See also [How do I create datacubes from a directory of GeoTIFFs](#create-datacubes-from-a-directory-of-geotiffs) |
| 53 | +above. |
| 54 | + |
| 55 | +## ... dynamically update global metadata attributes |
| 56 | + |
| 57 | +Refer to section about [target attributes](guide.md#attributes) |
| 58 | +in the user guide. |
| 59 | + |
| 60 | +## ... find out what is limiting the performance |
| 61 | + |
| 62 | +Use the [logging](guide.md#logging) configuration see which processing steps |
| 63 | +use most of the time. |
| 64 | +Use the [profiling](guide.md#profiling) configuration to inspect in more |
| 65 | +detail which parts of the processing are the bottlenecks. |
| 66 | + |
| 67 | +## ... write a log file |
| 68 | + |
| 69 | +Use the following [logging](guide.md#logging) configuration: |
| 70 | + |
| 71 | +```json |
| 72 | +{ |
| 73 | + "logging": { |
| 74 | + "version": 1, |
| 75 | + "formatters": { |
| 76 | + "normal": { |
| 77 | + "format": "%(asctime)s %(levelname)s %(message)s", |
| 78 | + "style": "%" |
| 79 | + } |
| 80 | + }, |
| 81 | + "handlers": { |
| 82 | + "console": { |
| 83 | + "class": "logging.StreamHandler", |
| 84 | + "formatter": "normal" |
| 85 | + }, |
| 86 | + "file": { |
| 87 | + "class": "logging.FileHandler", |
| 88 | + "formatter": "normal", |
| 89 | + "filename": "zappend.log", |
| 90 | + "mode": "w", |
| 91 | + "encoding": "utf-8" |
| 92 | + } |
| 93 | + |
| 94 | + }, |
| 95 | + "loggers": { |
| 96 | + "zappend": { |
| 97 | + "level": "INFO", |
| 98 | + "handlers": ["console", "file"] |
| 99 | + } |
| 100 | + } |
| 101 | + } |
| 102 | +} |
| 103 | +``` |
| 104 | + |
| 105 | +## ... address common errors |
| 106 | + |
| 107 | +### Error `Target parent directory does not exist` |
| 108 | + |
| 109 | +For security reasons, `zappend` does not create target directories |
| 110 | +automatically. You should make sure the parent directory exists before |
| 111 | +calling `zappend`. |
| 112 | + |
| 113 | +### Error `Target is locked` |
| 114 | + |
| 115 | +In this case the target lock file still exists, which means that a former |
| 116 | +rollback did not complete nominally. You can no longer trust the integrity of |
| 117 | +any existing target dataset. The recommended way is to remove the lock file |
| 118 | +and any target datasets artifact. You can do that manually or use the |
| 119 | +configuration setting `force_new`. |
| 120 | + |
| 121 | +### Error `Append dimension 'foo' not found in dataset` |
| 122 | + |
| 123 | +Refer to [How do I create datacubes from datasets without append dimension](#create-datacubes-from-datasets-without-append-dimension). |
0 commit comments