JP6339989B2 - Quality standard calculation device and quality standard calculation method - Google Patents

Description

  The present invention relates to a technique for calculating a quality standard used for software quality inspection.

  It is extremely important to inspect the quality of the developed software, and at present, a technology for appropriately determining and setting the quality standard is required. As a software quality standard, a bug density indicating the number of bugs with respect to the development scale and the number of steps of the source code is usually used. The bug density is set as an index value of the software quality standard, the upper limit and the lower limit of the index value are determined, and the index range included in these limits is set in advance as the control limit range (= allowable range of quality inspection). . When new software is developed, the quality of the software is determined based on whether or not the bug density of the new software falls within the management limit range. Therefore, the index value and the control limit range are important elements for determining the quality of software quality, and are carefully determined and set based on the past development results or the experience of the quality inspector (FIG. 5, Patent Document 1, 2).

JP 2013-225227 A JP 2012-173829 A

  However, there are concerns about the validity of the index values and control limit ranges that are determined and set because they are often determined sensuously. Specifically, when the degree of similarity with past software is low, the accuracy of the calculated control limit range, etc. will be low, and it will not be possible to set the control limit range suitable for new software, so software quality inspection will be appropriate (See FIG. 6). Even if only some of the functions added to the past software have been changed, the accuracy of the management limit range, etc. will also deteriorate even if the control limit range etc. are determined including other functions. (See FIG. 6).

  The present invention has been made in view of the above circumstances, and an object thereof is to optimize the allowable range (= control limit range) of software quality inspection.

  In order to solve the above problems, the quality standard calculation apparatus according to claim 1 includes a calculation unit that calculates the similarity between newly developed software and software developed in the past based on the similarity of the source code. The bug density for software development scale is calculated by using the bug density of software developed in the past as an index value, calculating the upper limit value obtained by adding a predetermined value to the index value and the lower limit value obtained by subtracting it, and correcting by the similarity. The recording means for recording in the U control chart representing the density, and reading out the U control chart from the storage means to obtain the corrected upper limit value and lower limit value corresponding to the development scale of the newly developed software. And an output unit that outputs an index range included in the two values as an allowable range of quality inspection for the new software.

  According to the present invention, the similarity between newly developed software and software developed in the past is calculated based on the similarity of the source code, the bug density of the software developed in the past is used as the index value, and the index Software quality inspection in order to calculate an upper limit value obtained by adding a predetermined value to the value and a lower limit value obtained by subtraction, and to correct the upper limit value and lower limit value with the above-mentioned similarity (= similarity based on similarity of source code). The permissible range (= the index range included between the upper limit value and the lower limit value) can be optimized.

  3. The quality standard calculation apparatus according to claim 2, wherein the recording unit adds the index value to a value obtained by multiplying the predetermined value by an inverse of the similarity. The gist is to use the value obtained as the corrected upper limit value and to subtract the value multiplied by the index value as the corrected lower limit value.

  The quality standard calculation apparatus according to claim 3 is the quality standard calculation apparatus according to claim 1 or 2, wherein the calculation means is similar between functions of newly developed software and software developed in the past. The gist is to calculate the degree based on the similarity of the source code between functions.

  The quality standard calculation device according to claim 4 is the quality standard calculation device according to claim 3, wherein the calculation means calculates the similarity between the functions based on the Euclidean distance.

  The quality standard calculation apparatus according to claim 5 is the quality standard calculation apparatus according to claim 3 or 4, wherein the software is composed of a plurality of functions, and the calculation means includes each source code corresponding to each function. The configuration value is normalized between the functions, and the similarity between the functions is calculated using the normalized configuration value.

  The quality standard calculation apparatus according to claim 6 is the quality standard calculation apparatus according to claim 5, wherein the configuration value of the source code includes at least the number of files, the number of classes, the number of methods, and the declaration sentence related to the source code. The gist is either the number of execution lines, the number of comments, the total number of source lines, the maximum number of nestings, or the complexity.

  The quality standard calculation apparatus according to claim 7 is the quality standard calculation apparatus according to any one of claims 1 to 6, wherein the output means has been developed in the past with the development difficulty level of newly developed software. The gist is to correct the development scale of the newly developed software in proportion to the software development difficulty level.

  The quality standard calculation method according to claim 8 is a quality standard calculation method performed by the quality standard calculation device, and calculates the similarity between newly developed software and software developed in the past based on the similarity of the source code. And developing the software by using the bug density of the software developed in the past as an index value, calculating an upper limit value obtained by adding a predetermined value to the index value and a lower limit value obtained by subtracting the index value, and correcting by the similarity. A step of recording the U control chart representing the bug density with respect to the scale; and reading the U control chart from the storage means to obtain an upper limit value and a lower limit value after correction corresponding to the development scale of the newly developed software; And a step of outputting an index range included in the obtained two values as an allowable range of quality inspection for the new software.

  According to the present invention, the allowable range (= control limit range) of software quality inspection can be optimized.

It is a figure which shows the functional block structure of a quality reference | standard calculation apparatus. It is a figure which shows operation | movement of a quality standard calculation apparatus. It is a figure which shows the example of a data set, such as the number of files regarding a development function. It is a figure which shows the example of a production | generation of a basic U control chart. It is a figure which shows the example of a setting of an index value and a management limit range. It is a reference figure at the time of explaining the subject of a prior art.

  Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing a functional block configuration of a quality standard calculation apparatus 1 according to the present embodiment. The quality standard calculation device 1 includes a past development data storage unit 11, a new development data storage unit 12, a development function similarity calculation unit 13, a basic U control chart generation unit 14, a past development difficulty calculation unit 15, A new development difficulty level calculation unit 16, a development scale correction unit 17, and an index value / management limit range calculation unit 18 are configured.

  The past development data storage unit 11 stores analysis result data and program related data regarding each software developed in the past. Specifically, for each software developed in the past, (1) development scale (number of design lines, etc.), number of bugs, (2) man-hours (man-months), development time (months), total amount of functions (FP (Function Point), the total number of functions, etc.), the development skill level of the software developer, and (3) the number of files related to the source code of each function, the number of classes, etc. are stored.

  The newly developed data storage unit 12 stores analysis result data and program related data regarding newly developed software. Specifically, for newly developed software, (1) development scale, (2) man-hours, development time, total amount of functions, development skill level of software developer, (3) number of files related to the source code of each function , The number of classes, etc. are stored.

  The development function similarity calculation unit 13 uses the number of files, the number of classes, and the like related to the source code of each function relating to each previously developed software and the newly developed software and the previously developed software, The similarity between software and functions is calculated based on the similarity of source code.

  The basic U control chart generation unit 14 uses the bug density of software developed in the past as an index value, and uses the development scale and the number of bugs of each software developed in the past as a basic index value (= bug density) and basic management. After calculating the limit range (= upper limit / lower limit) and correcting the basic control limit range with the above similarity (= similarity based on the similarity of the source code), the bug density relative to the software development scale was expressed It draws on the U chart and stores it in the memory as a basic U chart.

  The past development difficulty level calculation unit 15 calculates the past software development difficulty level using the man-hours, development time, total amount of functions, and development skill level related to the software developed in the past.

  The new development difficulty level calculation unit 16 calculates the development difficulty level of new software using the man-hours, development time, total function amount, and development skill level for newly developed software.

  The development scale correction unit 17 corrects the development scale of the newly developed software based on the ratio between the past software development difficulty and the new software development difficulty.

  The index value / control limit range calculation unit 18 reads the basic U control chart from the memory, refers to the corrected basic control limit range, and corrects the basic control limit after correction corresponding to the corrected development scale related to the new software. The range (= upper limit value after correction / lower limit value after correction) is obtained, and the index range included in the obtained basic control limit range is output as an allowable range for quality inspection for new software.

  This is the function of the quality standard calculation device 1. In addition, this quality standard calculation apparatus 1 is realizable with the computer provided with CPU, memory, etc. It is also possible to create a quality standard calculation program for causing a computer to function as the quality standard calculation apparatus 1 and a storage medium for the quality standard calculation program.

  Next, the operation of the quality standard calculation apparatus 1 will be described with reference to FIGS. 1 and 2. FIG. 2 is a diagram illustrating a processing flow of the quality standard calculation apparatus 1.

  First, the development function similarity calculation unit 13 uses the number of files, the number of classes, and the like related to the source code of the development function to convert the similarity between the new development function and the past development function into the similarity of the source code between the development functions. Based on the calculation (step S1).

For example, the newly developed software and historical development software, file count and the like according to the source code of each development functions of FIG. 3 (a), F development capabilities, metrics metrics file such as the number M, in the development functions F _i The actual data of M _j (= configuration value of the source code) is V _{i, j} (see FIG. 3B).

First, the actual data V _{i, j} is normalized between the development functions F _i (step S1-1). This normalization actual data V _i recorded in the data _set, since _j is the value range differs greatly depending metrics M _j, performed in order to equalize the effect of the prediction for each metric M _j. Specifically, the actual data V after normalization using the formula (1) when the actual data with the maximum value of the metrics M _j is max (V _j ) and the actual data with the minimum value is min (V _j ). 'Calculate _{i, j} . This normalization is performed for both newly developed functions and past developed functions.

続いて、新規開発機能と過去開発機能との類似度を計算する（ステップＳ１−２）。例えば、ユークリッド距離に基づき開発機能間の類似度を計算する。具体的には、新規開発機能をＦ_ａ、過去開発機能をＦ_ｉ、ユークリッド距離をｄｉｓｔ、新規開発機能Ｆ_ａと過去開発機能Ｆ_ｉとの類似度をｓｉｍ（Ｆ_ａ，Ｆ_ｉ）としたとき、式（２）を用いて類似度を計算する。

Subsequently, the similarity between the newly developed function and the past developed function is calculated (step S1-2). For example, the similarity between the development functions is calculated based on the Euclidean distance. Specifically, the newly developed function is F _a , the past developed function is F _i , the Euclidean distance is dist, and the similarity between the newly developed function F _a and the past developed function F _i is sim (F _a , F _i ). The similarity is calculated using equation (2).

式（２）ではユークリッド距離の逆数を類似度としているので、開発機能間の類似度はユークリッド距離が小さいほど大きくなる。この式（２）より、新規開発機能と過去開発機能との類似度がソースコードの類似性に基づき計算されることになる。なお、ｙはメトリクスＭ_ｊの総数である。

In equation (2), since the reciprocal of the Euclidean distance is used as the similarity, the similarity between the development functions increases as the Euclidean distance decreases. From this equation (2), the similarity between the newly developed function and the past developed function is calculated based on the similarity of the source code. Note that y is the total number of metrics M _j .

Finally, the calculated similarity between the development functions is set as the similarity sim (S _a , S _i ) between the software (step S1-3). S _a is newly developed software, and S _i is past developed software. For example, if the number of past development functions is one, the degree of similarity between the calculated development functions is one. On the other hand, when the number of past development functions is plural, for example, the average value of the similarity is set as the similarity between software.

  Next, after step S1, the basic U chart generation unit 14 uses the development scale and the number of bugs at the time of past development and the similarity between the software and the development function obtained in step S1, and A basic control limit range (= upper limit value / lower limit value) is calculated, and a basic U control chart is generated (step S2).

Specifically, the development scale at the time of past development is n, the number of bugs at the time of past development is c, the basic index value is u (“-” above u), and the basic management limit range is u _l (above u) "-"), Where the coefficient is k, the basic index value (= bug density) is calculated using equation (3), and the basic control limit range (= upper limit / lower limit) using equation (4) Is calculated (step S2-1).

式（４）では、基礎指標値ｕに対して加算・減算される所定値（＝ｋ√（ｕ／ｎ_ｉ））に対して、ステップＳ１で求めたソフトウェア間・開発機能間の類似度の逆数が乗算されている。これにより、基礎管理限界範囲（＝上限値・下限値）が該類似度（すなわち、ソースコードの類似性に基づく類似度）に基づき調整（補正）され、適正化されることになる（図４の補正Ａ参照）。

In Expression (4), the degree of similarity between the software and the development function obtained in step S1 is calculated with respect to a predetermined value (= k√ (u / n _i )) that is added to or subtracted from the basic index value u. The reciprocal is multiplied. As a result, the basic management limit range (= upper limit value / lower limit value) is adjusted (corrected) based on the similarity (that is, the similarity based on the similarity of the source code) and is optimized (FIG. 4). Correction A).

  After that, the basic control limit range after correction (= upper limit value after correction / lower limit value after correction) is plotted on the U control chart with the development scale on the horizontal axis and the bug density on the vertical axis. (See FIG. 4) is generated (step S2-2). This basic U chart is used later to determine the acceptable range (= control limit range) of quality inspection for newly developed software.

  Note that k is a coefficient for determining the upper and lower limits, and is arbitrarily set according to the development content. The past development scale n is, for example, the number of design digits. For the definition of U control chart, index value, control limit, etc., Japanese Industrial Standard JIS Z9020-1: 2011 (ISO 7870-1: 2007) (http://kikakurui.com/z9/Z9020-1-2011- 01.html), detailed in JIS Z9021: 1998 (http://kikakurui.com/z9/Z9021-1998-01.html) of the same standard.

Next, after step S2, the past development difficulty level calculation unit 15 obtains the man-hours, development time, total amount of functions, and development skill level related to the past development software from the past development data storage unit 11, and uses equation (5). calculating a historical development difficulty D _b of the past development software Te (step S3).

なお、βは定数である。過去開発ソフトウェアが複数の場合、各過去開発ソフトウェアの開発難易度の平均値を過去開発ソフトウェアの開発難易度としてもよい。

Note that β is a constant. When there are a plurality of past development software, the average value of the development difficulty of each past development software may be set as the development difficulty of the past development software.

Next, after step S3 or simultaneously with step S3, the new development difficulty level calculation unit 16 obtains the man-hour, development time, total function amount, and development skill level related to the newly developed software from the new development data storage unit 12, and formula calculating a new development difficulty D _a newly developed software with (6) (step S4).

次に、ステップＳ３及びステップＳ４の後、開発規模補正部１７が、式（７）を用いて、過去開発難易度Ｄ_ｂと新規開発難易度Ｄ_ａの比で新規開発ソフトウェアの開発規模ｎ_ａを補正する（ステップＳ５）。具体的には、新規開発ソフトウェアの開発規模ｎ_ａに対して、新規開発難易度Ｄ_ａを過去開発難易度Ｄ_ｂで除した値を乗算し、その乗算結果を新規開発ソフトウェアの補正後の開発規模ｎ’_ａとする。その際、比を計算するのに代えて差分を計算してもよい。

Then, after step S3 and step S4, development scale correction unit 17, using equation (7), development scale n _a newly developed software in a ratio of past development difficulty D _b and new development difficulty D _a Is corrected (step S5). Specifically, with respect to development scale n _a newly developed software, by multiplying the value obtained by dividing the new development difficulty D _a past development difficulty D _b, the development of the corrected new development software the multiplication result Let n ′ _a be the scale. At that time, the difference may be calculated instead of calculating the ratio.

式（７）より、新規開発ソフトウェアの開発規模ｎ_ａに対して過去開発難易度Ｄ_ｂと新規開発難易度Ｄ_ａの比を乗算するので、新規開発ソフトウェアの開発規模ｎ_ａは、過去開発と新規開発の両方の開発難易度に基づき調整（補正）され、適正化されることになる（図４の補正Ｂ参照）。

From the equation (7), since the multiplying ratio of the past development difficulty D _b and new development difficulty D _a relative development scale n _a newly developed software, development scale n _a newly developed software, past development and Adjustment (correction) is made based on the development difficulty levels of both newly developed and optimized (see correction B in FIG. 4).

Finally, after step S5, the index value / control limit calculation unit 18 acquires the basic U control chart of FIG. 4 generated in step S2 from the memory, and the development scale n ′ _a after correcting the newly developed software The basic control limit range (= upper limit value after correction / lower limit value after correction) is calculated, and the index range included in the calculated basic control limit range is output as an allowable range for quality inspection for new software (step) S6).

  As described above, according to the present embodiment, the development function similarity calculation unit 13 calculates the similarity between newly developed software and software developed in the past based on the similarity of the source code. The control chart generation unit 14 uses the bug density of software developed in the past as an index value, calculates an upper limit value obtained by adding a predetermined value to the index value, and a lower limit value obtained by subtracting the upper limit value and the lower limit value. Since the correction is performed with the similarity (= similarity based on the similarity of the source code), the allowable range of software quality inspection (= the index range included between the upper limit value and the lower limit value) can be optimized.

  In addition, according to the present embodiment, the development scale correction unit 17 compares the development difficulty level of newly developed software with the development difficulty level of previously developed software. Since the development scale is corrected, the development scale of the new software can be optimized.

  Finally, the data set in FIG. 3A is merely an example. Other metrics may be used. In the case of FIG. 3 (a), at least the number of files, the number of classes, the number of methods, the number of declaration statements and the number of executions, the number of comments, the total number of source lines, the maximum number of nestings, the complexity, Any one of them may be used.

  The U chart is a chart for managing the bug density in accordance with the development scale of software. In the case of software management, it is generally created based on the idea that the probability of deviating from the control limit with respect to the average value or index value of the bug density is about 0.3% (= coefficient k).

DESCRIPTION OF SYMBOLS 1 ... Quality standard calculation apparatus 11 ... Past development data storage part 12 ... New development data storage part 13 ... Development function similarity calculation part (calculation means)
14 ... Basic U control chart generator (recording means)
DESCRIPTION OF SYMBOLS 15 ... Past development difficulty calculation part 16 ... New development difficulty calculation part 17 ... Development scale correction | amendment part 18 ... Index value and management limit range calculation part (output means)
S1-S6 ... Step

Claims (8)

A calculation means for calculating the similarity between newly developed software and software developed in the past based on the similarity of the source code;
Using the bug density of software developed in the past as an index value, the upper limit value obtained by adding a predetermined value to the index value and the lower limit value obtained by subtraction are calculated, corrected by the similarity, and the bug density with respect to the software development scale Recording means for recording in a U chart representing
The U control chart is read from the storage means, the corrected upper limit value and lower limit value corresponding to the development scale of the newly developed software are obtained, and the index range included in the obtained two values is determined for the new software Output means for outputting as an acceptable range for quality inspection;
A quality standard calculation device comprising: The recording means includes
A value obtained by adding the index value to a value obtained by multiplying the reciprocal of the similarity by the predetermined value is set as an upper limit value after correction, and a value obtained by subtracting the multiplied value from the index value is set as a corrected lower limit value. The quality standard calculation apparatus according to claim 1, wherein: The calculating means includes
The quality standard calculation apparatus according to claim 1 or 2, wherein a similarity between functions of newly developed software and software developed in the past is calculated based on similarity of source codes between the functions. . The calculating means includes
The quality standard calculation apparatus according to claim 3, wherein the similarity between the functions is calculated based on a Euclidean distance. The software consists of multiple functions,
The calculating means includes
5. The configuration value of each source code corresponding to each function is normalized between functions, and the similarity between the functions is calculated using the normalized configuration value. Quality standard calculation device. The configuration value of the source code is
At least the number of files, classes, methods, declaration statements and execution lines, comments, source total lines, maximum nesting, complexity regarding source code The quality standard calculation apparatus according to claim 5. The output means includes
7. The development scale of the newly developed software is corrected based on a ratio between the development difficulty of newly developed software and the development difficulty of software developed in the past. The quality standard calculation device according to any one of the above. In the quality standard calculation method performed by the quality standard calculation device,
Calculating the similarity between newly developed software and previously developed software based on the similarity of the source code;
Using the bug density of software developed in the past as an index value, the upper limit value obtained by adding a predetermined value to the index value and the lower limit value obtained by subtraction are calculated, corrected by the similarity, and the bug density with respect to the software development scale Recording in a U chart representing
The U control chart is read from the storage means, the corrected upper limit value and lower limit value corresponding to the development scale of the newly developed software are obtained, and the index range included in the obtained two values is determined for the new software Outputting as inspection inspection tolerance,
A quality standard calculation method comprising:

Images