用jq分解復雜的JSON物件（帶有物件和陣列）

我們正在接收來自客戶的資料并且它是高度嵌套的（客戶這樣做是為了減少重復資料的數量，從而減少傳輸資料的整體大小）。

資料如下所示：

{
    "r1": "ex",
    "r2": "of",
    "r3": "da",
    "ver": 1,
    "noa1": [{
        "col1": 380,
        "noa2": [{
            "aon": [123],
            "obj": {
                "col10": "stuf",
                "col11": "here",
                "aos": ["A",
                    "X"
                ]
            }
        }, {
            "aon": [],
            "obj": {
                "col10": "more",
                "col11": "stuf"
            }
        }, {
            "aon": [456, 789]
        }, {
            "obj": {
                "col10": "anon"
            }
        }]
    }, {
        "col1": 5676,
        "noa2": [{
            "aon": [875],
            "obj": {
                "col10": "does",
                "col11": "noth",
                "aos": ["Y",
                    "Z"
                ]
            }
        }, {
            "obj": {
                "col11": "phew"
            }
        }]
    }]
}

其中有許多可選的嵌套物件和陣列（大約 6 級深 IRL，但在此處進行了簡化）。

我們必須將其展平為 CSV，其中每一行都是唯一的，其中包含 JSON 物件的資料。我們不希望從跨積連接創建任何“新行”。所以上面看起來像：

r1|r2|r3|ver|col1|aon|col10|col11|aos
ex|of|da|1  |380 |123|stuf |here |A
ex|of|da|1  |380 |123|stuf |here |X
ex|of|da|1  |380 |   |more |stuf |
ex|of|da|1  |380 |456|     |     |
ex|of|da|1  |380 |789|     |     |
ex|of|da|1  |380 |   | anon|     |
ex|of|da|1  |5676|875|does |noth |Y
ex|of|da|1  |5676|875|does |noth |Z
ex|of|da|1  |5676|   |     |phew |

For brevity, we show this as a table. In reality, we would want this to be a collection of JSON documents (jq -c).

我們一直在使用 jq，但在使用不是物件的嵌套陣列時遇到了困難，比如字串陣列。它導致發生交叉連接，將值放在原始 JSON 中不存在的位置。我們也遇到了可選嵌套行的問題（如果陣列為空，jq 當前不會列印該行，即使使用 ? 修飾符也是如此。

問題是：

1. 原始 JSON 的 jq 過濾器是什么？
2. 一般來說，如何處理 jq 中的嵌套物件、陣列以及字串、數字等陣列？

更新：以下代碼

.r1 as $r1 | .r2 as $r2 | .r3 as $r3 | .ver as $ver | .noa1[] | . as $noa1 | .noa2[] | . as $noa2 | {$r1, $r2, $r3, $ver, "col1": $noa1.col1, "aon": $noa2.aon }

回傳

{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":380,"aon":[123]}
{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":380,"aon":[]}
{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":380,"aon":[456,789]}
{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":380,"aon":null}
{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":5676,"aon":[875]}
{"r1":"ex","r2":"of","r3":"da","ver":1,"col1":5676,"aon":null}

這是有道理的。但是現在，當我們嘗試更進一步并分解“aon”數字陣列時，它變得很奇怪。這段代碼

.r1 as $r1 | .r2 as $r2 | .r3 as $r3 | .ver as $ver | .noa1[] | . as $noa1 | .noa2[] | . as $noa2 | 
{$r1, $r2, $r3, $ver, "col1": $noa1.col1, "aon": $noa2.aon[]? }

丟失具有“aon”空值的行，我可以理解，有點（因為？）。但是這段代碼失敗了

.r1 as $r1 | .r2 as $r2 | .r3 as $r3 | .ver as $ver | .noa1[] | . as $noa1 | .noa2[] | . as $noa2 | 
{$r1, $r2, $r3, $ver, "col1": $noa1.col1, "aon": $noa2.aon[] }

我也明白，因為空值。所以我們嘗試將它作為一個變數來參考：

.r1 as $r1 
| .r2 as $r2 
| .r3 as $r3 
| .ver as $ver 
| .noa1[] | . as $noa1 
| .noa2[] | . as $noa2
| .obj as $obj
| .aon[]? | . as $aon
| {$r1, $r2, $r3, $ver, "col1": $noa1.col1, "aon": $aon, "col10": $obj.col10, "col11": $obj.col11 }

并走到這一步。但是有幾個問題：我們丟失了一些 aon 為空的行，我們不知道如何訪問“aos”。

uj5u.com熱心網友回復：

我不確定您是否希望“aos”欄位出現在每個生成的物件中。如果沒有，您可能會做得比：

jq -c '
  .r1 as $r1
  | .r2 as $r2
  | .r3 as $r3
  | .ver as $ver
  | .noa1[] 
  | .col1 as $col1
  | .noa2[] 
  | .aon as $aon
  | .obj
  | {$r1, $r2, $ver, $col1, $aon, col10, col11} 
      try {aos: .aos[]} catch {} '

如果你這樣做，那么你可以將最后一個替換為{}或{aos: null}其他。如果您不想使用，try ... catch那么您可以用一個if陳述句替換它，或者可以用：

( (.aos // [null]) | {aos: .[]} )

我認為，保持理智的關鍵是避免創建 jq 變數，除非“深入”時需要。此外，jq 對 {$x} 和 {y} 等縮寫的支持使事情更容易閱讀和審查。

標籤：jsonUnixjq

上一篇：替換部分fasta標頭

下一篇：不懂Linux命令