exp show --csv: Parent column empty

[dberenbaum]

Bug Report

Description

The parent commit column in exp show --csv is empty even for checkpoints where a parent commit exists.

Reproduce

$ git clone git@github.com:dberenbaum/dvc-checkpoint.git
$ echo 1 > bar
$ dvc exp run
$ dvc exp show --json
{
  "workspace": {
    "baseline": {
      "data": {
        "timestamp": null,
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 1,
              "mult": 1
            }
          }
        }
      }
    }
  },
  "0fd92c2c0ff17c923a79b4b0b96e54fe4b73a8cf": {
    "baseline": {
      "data": {
        "timestamp": "2021-03-10T14:05:52",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {},
        "name": "main"
      }
    },
    "0a2b7379f2720a2fe240a826a611c285ddb2d07f": {
      "data": {
        "checkpoint_tip": "0a2b7379f2720a2fe240a826a611c285ddb2d07f",
        "timestamp": "2021-11-18T20:30:27",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 1,
              "mult": 1
            }
          }
        },
        "name": "exp-e44da",
        "checkpoint_parent": "8acf8ec4ea2dcedc39eac011d35cf26a1e1f291d"
      }
    },
    "8acf8ec4ea2dcedc39eac011d35cf26a1e1f291d": {
      "data": {
        "checkpoint_tip": "0a2b7379f2720a2fe240a826a611c285ddb2d07f",
        "timestamp": "2021-11-18T20:30:26",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 0,
              "mult": 0
            }
          }
        },
        "checkpoint_parent": "0fd92c2c0ff17c923a79b4b0b96e54fe4b73a8cf"
      }
    }
  }
}
$ dvc exp show --csv
Experiment,rev,typ,Created,parent,epoch,mult,start,EPOCHS,STEP
,workspace,baseline,,,1,1,0,2,1
main,0fd92c2,baseline,2021-03-10T14:05:52,,,,0,2,1
exp-e44da,0a2b737,checkpoint_tip,2021-11-18T20:30:27,,1,1,0,2,1
,8acf8ec,checkpoint_base,2021-11-18T20:30:26,,0,0,0,2,1

Expected

8acf8ec should have 0fd92c2c0ff17c923a79b4b0b96e54fe4b73a8cf in parent.

Environment information

2.8.4.dev29+gd8c77da4

[pmrowla]

I'm not sure this is a bug, from the related PR:

The checkpoint_parent column is only intended to be filled in the "checkpoint branch off of an existing checkpoint commit" scenario. In the typical use case where a checkpoint branches off of a (non-DVC experiment) parent git commit, checkpoint_parent is left empty, since it is always just the same as the regular experiment "baseline" git commit.

[dberenbaum]

Sorry, this was not well explained at all, and I was also confused. Here's the example again with the rich table results to make it easier to see what's happening:

$ git clone git@github.com:dberenbaum/dvc-checkpoint.git

$ echo 1 > bar

$ dvc exp run
 ─────────────────────────────────────────────────────────────────────────────────────────────────────────
  Experiment                Created        epoch   mult   start   EPOCHS   STEP   bar       checkpoint.py
 ─────────────────────────────────────────────────────────────────────────────────────────────────────────
  workspace                 -                  1      1   0       2        1      b026324   443a8a8
  main                      Mar 10, 2021       -      -   0       2        1      -         -
  │ ╓ fe9cf04 [exp-e44da]   08:41 AM           1      1   0       2        1      b026324   443a8a8
  ├─╨ 7d5a9a4               08:41 AM           0      0   0       2        1      b026324   443a8a8
 ─────────────────────────────────────────────────────────────────────────────────────────────────────────

$ dvc exp show --csv
Experiment,rev,typ,Created,parent,epoch,mult,start,EPOCHS,STEP,bar,checkpoint.py
,workspace,baseline,,,1,1,0,2,1,b026324,443a8a8
main,0fd92c2,baseline,2021-03-10T14:05:52,,,,0,2,1,,
exp-e44da,fe9cf04,checkpoint_tip,2022-01-27T08:41:36,,1,1,0,2,1,b026324,443a8a8
,7d5a9a4,checkpoint_base,2022-01-27T08:41:35,,0,0,0,2,1,b026324,443a8a8

$ dvc exp show --json
{
  "workspace": {
    "baseline": {
      "data": {
        "timestamp": null,
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "deps": {
          "checkpoint.py": {
            "hash": "443a8a88f4096afe54ee4909a09b5ecb",
            "size": 1204,
            "nfiles": null
          },
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          }
        },
        "outs": {
          "foo": {
            "hash": "c4ca4238a0b923820dcc509a6f75849b",
            "size": 1,
            "nfiles": null
          },
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 1,
              "mult": 1
            }
          }
        }
      }
    }
  },
  "0fd92c2c0ff17c923a79b4b0b96e54fe4b73a8cf": {
    "baseline": {
      "data": {
        "timestamp": "2021-03-10T14:05:52",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "deps": {
          "checkpoint.py": {
            "hash": null,
            "size": null,
            "nfiles": null
          },
          "bar": {
            "hash": null,
            "size": null,
            "nfiles": null
          }
        },
        "outs": {
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          },
          "foo": {
            "hash": null,
            "size": null,
            "nfiles": null
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {},
        "name": "main"
      }
    },
    "fe9cf04818a076e1738450a13e7a573fc39110a3": {
      "data": {
        "checkpoint_tip": "fe9cf04818a076e1738450a13e7a573fc39110a3",
        "timestamp": "2022-01-27T08:41:36",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "deps": {
          "checkpoint.py": {
            "hash": "443a8a88f4096afe54ee4909a09b5ecb",
            "size": 1204,
            "nfiles": null
          },
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          }
        },
        "outs": {
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          },
          "foo": {
            "hash": "c4ca4238a0b923820dcc509a6f75849b",
            "size": 1,
            "nfiles": null
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 1,
              "mult": 1
            }
          }
        },
        "name": "exp-e44da",
        "checkpoint_parent": "7d5a9a4de8ef5f94e5ef98c436cec6de8a05ec8a"
      }
    },
    "7d5a9a4de8ef5f94e5ef98c436cec6de8a05ec8a": {
      "data": {
        "checkpoint_tip": "fe9cf04818a076e1738450a13e7a573fc39110a3",
        "timestamp": "2022-01-27T08:41:35",
        "params": {
          "params.yaml": {
            "data": {
              "start": 0
            }
          },
          "checkpoint.py": {
            "data": {
              "EPOCHS": 2,
              "STEP": 1
            }
          }
        },
        "deps": {
          "checkpoint.py": {
            "hash": "443a8a88f4096afe54ee4909a09b5ecb",
            "size": 1204,
            "nfiles": null
          },
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          }
        },
        "outs": {
          "bar": {
            "hash": "b026324c6904b2a9cb4b88d6d61c81d1",
            "size": 2,
            "nfiles": null
          },
          "foo": {
            "hash": "cfcd208495d565ef66e7dff9f98764da",
            "size": 1,
            "nfiles": null
          }
        },
        "queued": false,
        "running": false,
        "executor": null,
        "metrics": {
          "scores.yaml": {
            "data": {
              "epoch": 0,
              "mult": 0
            }
          }
        },
        "checkpoint_parent": "0fd92c2c0ff17c923a79b4b0b96e54fe4b73a8cf"
      }
    }
  }
}

In the rich table, current behavior makes sense since there is a visual cue of how the checkpoints connect to their parent. In the CSV, I think it makes the most sense to duplicate experiment info as much as possible in each checkpoint so that you don't need to rely on the row order or looking up experiment info from another row (including as much info as possible in the CSV was also discussed in #6848). At least, the CSV should include the experiment name for each checkpoint:

$ dvc exp show --csv
Experiment,rev,typ,Created,parent,epoch,mult,start,EPOCHS,STEP,bar,checkpoint.py
,workspace,baseline,,,1,1,0,2,1,b026324,443a8a8
main,0fd92c2,baseline,2021-03-10T14:05:52,,,,0,2,1,,
exp-e44da,fe9cf04,checkpoint_tip,2022-01-27T08:41:36,,1,1,0,2,1,b026324,443a8a8
exp-e44da,7d5a9a4,checkpoint_base,2022-01-27T08:41:35,,0,0,0,2,1,b026324,443a8a8

If it's not too difficult, it would be nice to also include tip, baseline, and parent for each checkpoint:

$ dvc exp show --csv
Experiment,rev,typ,Created,parent,tip,baseline,epoch,mult,start,EPOCHS,STEP,bar,checkpoint.py
,workspace,baseline,,,,,1,1,0,2,1,b026324,443a8a8
main,0fd92c2,baseline,2021-03-10T14:05:52,,,0fd9sc2,,,0,2,1,,
exp-e44da,fe9cf04,checkpoint_tip,2022-01-27T08:41:36,7d5a9a4,fe9cf04,0fd92c2,,1,1,0,2,1,b026324,443a8a8
exp-e44da,7d5a9a4,checkpoint_base,2022-01-27T08:41:35,0fd92c2,fe9cf04,0fd92c2,,0,0,0,2,1,b026324,443a8a8

[dberenbaum]

@karajan1001 What's the status of this issue and #7004 ?