Android Performance Patterns - s4ep5 Minimizing Asset Payload

No matter how aggressively you’ve optimized your networking code for performance, there’s still a huge problematic performance issue that you haven’t seen: Bloated asset files. If you’re not aggressively attacking the size of the assets you’re transferring to your users, you’re basically stealing time (and money) from them.
Video Link

---

Smaller Is Better

1. 速度缓慢并不是所有用户都处于高速的4G／wifi环境下
2. 大数据量导致的大电量消耗
3. 流量＝金钱

引起数据流量过大的，主要有两个因素：图片、序列化数据。

---

Images

如果不需要图片透明度设定，那就尽量不要使用PNG格式的图片文件，这会产生难以置信的图片体积。

JPG和WEBP是不错的选择，在使用这两种图片格式时，记住图片质量的极小降低，可以大大缩小图片体积。

图片的分辨率并不是越大越好，根据具体的使用场景，返回不同像素的图片。如果只是需要缩略图，那就不要返回高清的无码大图。

---

Serialized Data

JSON /dʒeɪ’sʌn/ 是一个horrible的数据格式！！！

看看下面的代码，包含了大量的空格、引号、key值等冗余信息，这些字段存在的意义只是为了方便人工阅读，并非数据传输。

{
  "__name": "MOPayOrderDo",
  "RefundDetails": ["10.1元   2016-02-12 20:00 退款", "10.02元   2016-02-17 19:00 退款", "10.003元   2016-02-18 17:00 退款", "10.0004元   2016-02-18 18:00 退款"],
  "BaseOrderId": null,
  "OriAmount#D": 2.123456,
  "RemainAmount#D": 100.123456,
  "SerializedId": null,
  "RepayUrl": "dianping://shopinfo?id=2062015",
  "CurrentAmount#D": 0.01,
  "SaveAmount#D": 0,
  "OrderRemarks": null,
  "ErrorMsg": "支付成功|此为付款凭证，请向商户展示",
  "SerialNumber": "80033630111",
  "RightDos": null,
  "MerchantAmountString": "",
  "Discounts": null,
  "HuiTicketShareDo": {
      "__name": "HuiTicketShareDo",
      "TicketValidPeriod": "",
      "TicketDesc": "",
      "TicketValue": "",
      "TicketShareStatus#I": 0,
      "TicketButtonClickable": false,
      "PayTicketDesc": "",
      "TicketPicUrl": "",
      "TicketTitle": "",
      "TicketButtonText": "",
      "ShareDo": null,
      "ExtraTickets": ""
  },
  "ShopID#I": 2062015,
  "PointMallDo": null,
  "RefundAmount#D": 40.1234,
  "BuffetDescs": null,
  "PayFailDescription": "您的付款将于1-5个工作日原路退还",
  "Time#U": 1442469357,
  "HuiReviewInfo": null,
  "Banner": null,
  "OrderID#I": 43956789,
  "VoucherSerials": [],
  "Status#I": 3,
  "OperationBanners": null,
  "AlertLoginTips": "绑定手机后订单不会丢哦",
  "AlertLoginLink": {
      "__name": "Link",
      "Name": "现在去绑定",
      "Url": ""
  },
  "MerchantAmount#D": 0.01,
  "ServiceUrl": null,
  "StatusMsg": "已退款",
  "ContactMerchantTip": "",
  "OrderDetailTipsDo": null,
  "ShopPhones": [
      "4008205527"
  ],
  "ShopName": "大众点评网",
  "IsHobbit#I": 0,
  "NoDiscountAmount#D": 0,
  "SuccessMsg": "请向服务员出示验证码",
  "HasVoiceReport": false,
  "MobileNo": "13774283697",
  "VerifyStatus#I": 2,
  "MOPayShare": {
      "__name": "Share",
      "IconUrl": "http://qcloud.dpfile.com/pc/9vKBXqRXPl6vyOVFe4zlXMLys21f38u--LZmWbufGI2pRKzSBzQnLPwzWURFZUAvtOnd3gXQdDYlAqlaVaAFeZ0rYYyiRo_EhzufqWWjTjs.jpg",
      "Title": "在【大众点评网】不用现金，手机买单有优惠!",
      "Desc": "",
      "BtnText": null,
      "Url": "http://m.dianping.com/hui/share/weixin?shopId=2062015"
  },
  "UserAmountString": "{\"userAmount\":\"1.123456元\",\"boughtVoucher\":\"\",\"usedVoucher\":\"\"}",
  "BizType#I": 10
}

光头哥在video中给出了几种高效的序列化替代方式，讲解参见Serialization performance (Android Performance Patterns Season 4 ep14)

• Protocol Buffers
• Nano-Proto-Buffers
• FlatBuffers

同样，在http压缩过程里，也有进一步优化的空间，参见Smaller Serialized Data (Android Performance Patterns Season 4 ep15)

---

===Ending===

---